Code

merge from upstream#25

Code

Merged

l3utterfly merged 23 commits intolayla-buildl3utterfly/llama.cpp:layla-buildfrom masterl3utterfly/llama.cpp:masterCopy head branch name to clipboardJun 25, 2024

+24,175-23,045

Commits

Commits on Jun 21, 2024

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

  Show description for 557b653
  Adriankhl

  authored
  557b653

  Copy full SHA for 557b653

Commits on Jun 22, 2024

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

  Show description for c5a8d4b
  HanClinto

  authored
  c5a8d4b

  Copy full SHA for c5a8d4b

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

  Show description for 5b48cd5
  ddh0

  authored
  5b48cd5

  Copy full SHA for 5b48cd5

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

  0xspringtime

  authored
  3aa184a

  Copy full SHA for 3aa184a

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

  Show description for adf480c
  HatsuneMikuUwU33

  authored
  adf480c

  Copy full SHA for adf480c

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

  Show description for 3e58b0e
  ngxson

  authored
  3e58b0e

  Copy full SHA for 3e58b0e

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

  HanClinto

  authored
  b5a5f34

  Copy full SHA for b5a5f34

Commits on Jun 23, 2024

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

  Show description for 45c0e2e
  0cc4m

  authored
  45c0e2e

  Copy full SHA for 45c0e2e

• fix CI failures (ggerganov#8066)

  Show description for b6b9a8e
  slaren

  authored
  b6b9a8e

  Copy full SHA for b6b9a8e

• Fix typo in llama_set_embeddings comment (ggerganov#8077)

  danbev

  authored
  11318d9

  Copy full SHA for 11318d9

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

  akx

  authored
  6a2f298

  Copy full SHA for 6a2f298

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

  Show description for e112b61
  

  Eddie-Wang1120

  and

  root

  authored
  e112b61

  Copy full SHA for e112b61

Commits on Jun 24, 2024

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

  Show description for 95f57bb
  slaren

  authored
  95f57bb

  Copy full SHA for 95f57bb

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

  Show description for de0d6a6
  

  fairydreaming

  and

  sszymczy

  authored
  de0d6a6

  Copy full SHA for de0d6a6

• embedding : more cli arguments (ggerganov#7458)

  Show description for 646ef4a
  

  YannFollet

  and

  mofosyne

  authored
  646ef4a

  Copy full SHA for 646ef4a

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

  slaren

  authored
  8cb508d

  Copy full SHA for 8cb508d

• Option to split during conversion (ggerganov#6942)

  Show description for 52fc870
  

  3 peopleauthored
  52fc870

  Copy full SHA for 52fc870

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

  Show description for 9a590c8
  JohannesGaessler

  authored
  9a590c8

  Copy full SHA for 9a590c8

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

  Show description for d62e4aa
  

  3 peopleauthored
  d62e4aa

  Copy full SHA for d62e4aa

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

  JohannesGaessler

  authored
  a818f30

  Copy full SHA for a818f30

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

  JohannesGaessler

  authored
  3b099bc

  Copy full SHA for 3b099bc

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

  JohannesGaessler

  authored
  2df373a

  Copy full SHA for 2df373a

Commits on Jun 25, 2024

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

  airMeng

  authored
  083bacc

  Copy full SHA for 083bacc