Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Jun 25, 2024

• fixes ggerganov#7999 (adds control vectors to all build_XXX() functions in llama.cpp [needs testing] (ggerganov#8060)

  jukofyork

  authored
  163d50a

• llama : implement Unigram tokenizer needed by T5 and FLAN-T5 model families (ggerganov#5763)

  

  fairydreaming

  and

  sszymczy

  authored
  6fcbf68

• llama : return nullptr from llama_grammar_init (ggerganov#8093)

  

  danbev

  and

  HanClinto

  authored
  e6bf007

• json: support integer minimum, maximum, exclusiveMinimum, exclusiveMaximum (ggerganov#7797)

  ochafik

  authored
  84631fe

• disable docker CI on pull requests (ggerganov#8110)

  slaren

  authored
  dd047b4

• Add healthchecks to llama-server containers (ggerganov#8081)

  codearranger

  authored
  925c309

• Gguf dump start data offset via --data-offset and some extra refactor (ggerganov#8054)

  mofosyne

  authored
  c8ad359

• cvector: better prompt handling, add "mean vector" method (ggerganov#8069)

  ngxson

  authored
  49c03c7

• Add chat template support for llama-cli (ggerganov#8068)

  

  ngxson

  and

  ggerganov

  authored
  48e6b92

• SimpleChat v3.1: Boolean chat request options in Settings UI, cache_prompt (ggerganov#7950)

  hanishkvc

  authored
  3791ad2

• Update control vector help (ggerganov#8104)

  HatsuneMikuUwU33

  authored
  f702a90

• [SYCL] Re-enabled mul_mat_batched_sycl (ggerganov#8095)

  airMeng

  authored
  083bacc

Commits on Jun 24, 2024

• CUDA: fix matrix multiplication algorithm choice (ggerganov#8102)

  JohannesGaessler

  authored
  2df373a

• CUDA: fix MMQ writeback for int8 tensor cores (ggerganov#8100)

  JohannesGaessler

  authored
  3b099bc

• CUDA: use MMQ instead of cuBLAS by default (ggerganov#8075)

  JohannesGaessler

  authored
  a818f30

• gguf-py : fix tensor groups for encoder-decoder models in gguf-dump.py (ggerganov#8090)

  

  3 peopleauthored
  d62e4aa

• CUDA: optimize MMQ int8 tensor core performance (ggerganov#8062)

  JohannesGaessler

  authored
  9a590c8

• Option to split during conversion (ggerganov#6942)

  

  3 peopleauthored
  52fc870

• disable publishing the full-rocm docker image (ggerganov#8083)

  slaren

  authored
  8cb508d

• embedding : more cli arguments (ggerganov#7458)

  

  YannFollet

  and

  mofosyne

  authored
  646ef4a

• gguf-py, convert-hf : model conversion support for T5 and FLAN-T5 model variants (ggerganov#5763)

  

  fairydreaming

  and

  sszymczy

  authored
  de0d6a6

• ggml : remove ggml_task_type and GGML_PERF (ggerganov#8017)

  slaren

  authored
  95f57bb

Commits on Jun 23, 2024

• llama : add support for BitnetForCausalLM (ggerganov#7931)

  

  Eddie-Wang1120

  and

  root

  authored
  e112b61

• server : fix JSON-Scheme typo (ggerganov#7975)

  akx

  authored
  6a2f298

• Fix typo in llama_set_embeddings comment (ggerganov#8077)

  danbev

  authored
  11318d9

• fix CI failures (ggerganov#8066)

  slaren

  authored
  b6b9a8e

• Refactor Vulkan backend to allow multiple contexts (ggerganov#7961)

  0cc4m

  authored
  45c0e2e

Commits on Jun 22, 2024

• Removing extra blank lines that were breaking Lint. (ggerganov#8067)

  HanClinto

  authored
  b5a5f34

• cvector: fix CI + correct help message (ggerganov#8064)

  ngxson

  authored
  3e58b0e

• cvector-generator: Moe Moe Fixie-Fixie for Lots of Formats~! ♡(ᐢ ᴥ ᐢ)♡ (ggerganov#8052)

  HatsuneMikuUwU33

  authored
  adf480c

• convert-hf : change assert to exception (ggerganov#8015)

  0xspringtime

  authored
  3aa184a

• Update llama-quantize ppl/file size output from LLaMA-v1 to Llama-3 values (ggerganov#8058)

  ddh0

  authored
  5b48cd5

• JSON Schema to GBNF integration tests (ggerganov#7790)

  HanClinto

  authored
  c5a8d4b

Commits on Jun 21, 2024

• vulkan: detect multiple devices by deviceUUID instead of deviceID (ggerganov#8022)

  Adriankhl

  authored
  557b653

• ggml : AVX IQ quants (ggerganov#7845)

  netrunnereve

  authored
  7d5e877

Pagination

PreviousNext