Code

b2392#96

Code

Merged

Nexesenex merged 26 commits intoNexesenex:_master_upNexesenex/croco.cpp:_master_upfrom

ggml-org:masterggml-org/llama.cpp:masterCopy head branch name to clipboard

Mar 10, 2024

Commits

Commits on Mar 9, 2024

• server: tests: add truncated prompt tests, better kv cache size (#5933)

  Show description for fd72d2d
  

  phymbert

  and

  ggerganov

  authored
  fd72d2d

  Copy full SHA for fd72d2d

• server : add SSL support (#5926)

  Show description for e1fa956
  gabe-l-hart

  authored
  e1fa956

  Copy full SHA for e1fa956

• Server: reorganize some http logic (#5939)

  Show description for 950ba1a
  ngxson

  authored
  950ba1a

  Copy full SHA for 950ba1a

• server : simplify logic for empty prompts (#5953)

  ggerganov

  authored
  9674aaf

  Copy full SHA for 9674aaf

• ggml : add ggml-common.h to deduplicate shared code (#5940)

  Show description for 8a3012a
  ggerganov

  authored
  8a3012a

  Copy full SHA for 8a3012a

• server : fix passing prompt as tokens (#5955)

  Show description for 0db32be
  

  z80maniac

  and

  ggerganov

  authored
  0db32be

  Copy full SHA for 0db32be

• tests : gitignore ggml-common.h

  ggerganov

  committed
  2c4f566

  Copy full SHA for 2c4f566

• server : normalize embeddings (#5956)

  Show description for fb215c3
  

  redlion0929

  and

  ggerganov

  authored
  fb215c3

  Copy full SHA for fb215c3

• server : clarify some items in the readme (#5957)

  Show description for 97c0958
  ggerganov

  authored
  97c0958

  Copy full SHA for 97c0958

• ggml : remove old quantization functions (#5942)

  Show description for 5b09797
  ggerganov

  authored
  5b09797

  Copy full SHA for 5b09797

• server : fix metrics init (#5964)

  ggerganov

  authored
  58308a0

  Copy full SHA for 58308a0

• ggml : fix unnecessary f32 -> f16 -> f32 casts (mmla) (#5951)

  ggerganov

  authored
  8380ecf

  Copy full SHA for 8380ecf

• readme : update hot topics

  ggerganov

  authored
  098dbaa

  Copy full SHA for 098dbaa

• perplexity : support using multiple sequences to allow larger batch sizes (#5946)

  Show description for d894f35
  slaren

  authored
  d894f35

  Copy full SHA for d894f35

• server : print chat template info

  ggerganov

  committed
  77d1ac7

  Copy full SHA for 77d1ac7

• server: benchmark: chat/completions scenario and other llm servers comparison (#5941)

  Show description for 621e86b
  

  phymbert

  and

  ggerganov

  authored
  621e86b

  Copy full SHA for 621e86b

Commits on Mar 10, 2024

• nix: update flake.lock (#5969)

  Show description for c785414
  

  ggerganov

  and

  github-actions[bot]

  authored
  c785414

  Copy full SHA for c785414

• grammar : verify parsed state (#5950)

  HanClinto

  authored
  2960eae

  Copy full SHA for 2960eae

• llama : add support for GritLM (#5959)

  Show description for bcebd7d
  

  3 peopleauthored
  bcebd7d

  Copy full SHA for bcebd7d

• server: ci: windows build and tests (#5968)

  Show description for fa8a809
  

  phymbert

  and

  cebtenzzre

  authored
  fa8a809

  Copy full SHA for fa8a809

• ggml : remove __constant__ specifier for CUDA tables (#5940)

  ggerganov

  committed
  bf47a5e

  Copy full SHA for bf47a5e

• ggml : try fix 32-bit arm compat (whisper/1938)

  Show description for df4dc3e
  ggerganov

  committed
  df4dc3e

  Copy full SHA for df4dc3e

• sync : ggml

  ggerganov

  committed
  b838b53

  Copy full SHA for b838b53

• readme : update hot topics

  ggerganov

  authored
  d9f65c9

  Copy full SHA for d9f65c9

• android : fix utf8 decoding error (#5935)

  Show description for 7ab7b73
  

  3 peopleauthored
  7ab7b73

  Copy full SHA for 7ab7b73

• metal : move mm_id indices to shared mem (#5982)

  ggerganov

  authored
  bb6d00b

  Copy full SHA for bb6d00b