You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardexpand all lines: README.md
+1-1
Original file line number
Diff line number
Diff line change
@@ -25,7 +25,7 @@ ThunderKittens is built around three key principles:
25
25
<img src="assets/attn.png" height=600 alt="Flash Attention 3, but with kittens!" style="margin-bottom:px"/>
26
26
</div>
27
27
28
-
Join us on Discord to get involved: [ThunderKittens channel @ GPU Mode Discord](https://discord.com/channels/1189498204333543425/1300872762163728550)!!!! Here is the invite link to GPU mode: https://discord.gg/W7CUhN5w
28
+
Join us on Discord to get involved: [ThunderKittens channel @ GPU Mode Discord](https://discord.com/channels/1189498204333543425/1300872762163728550)!!!! Here is the invite link to GPU mode: https://discord.gg/gpumode
29
29
30
30
ThunderKittens is built from the hardware up -- we do what the silicon tells us. And modern GPUs tell us that they want to work with fairly small tiles of data. A GPU is not really a 1000x1000 matrix multiply machine (even if it is often used as such); it’s a manycore processor where each core can efficiently run ~16x16 matrix multiplies. Consequently, ThunderKittens is built around manipulating tiles of data no smaller than 16x16 values.
0 commit comments