Code

merge upstream#38

Code

Merged

l3utterfly merged 48 commits intolayla-buildl3utterfly/llama.cpp:layla-buildfrom mergel3utterfly/llama.cpp:mergeCopy head branch name to clipboardSep 23, 2024

+5,178-3,371

Commits

Commits on Sep 15, 2024

• cmake : correct order of sycl flags (ggerganov#9497)

  Xarbirus

  authored
  6988da9

• gguf-split : add basic checks (ggerganov#9499)

  Show description for e6deac3
  slaren

  authored
  e6deac3

• common : reimplement logging (ggerganov#9418)

  Show description for 6262d13
  ggerganov

  authored
  6262d13

Commits on Sep 16, 2024

• flake.lock: Update (ggerganov#9488)

  ggerganov

  authored
  90a2fff

• metal : handle zero-sized allocs (ggerganov#9466)

  ggerganov

  authored
  c4965a6

• main : option to disable context shift (ggerganov#9484)

  Show description for 441b72b
  

  VJHack

  and

  ggerganov

  authored
  441b72b

• llama : support MiniCPM3 (ggerganov#9322)

  Show description for 95ca851
  

  CarryFun

  and

  范睿凯

  authored
  95ca851

• llama : support OLMoE (ggerganov#9462)

  2015aroras

  authored
  0aadac1

• ggml : IQ4_NL sgemm + Q4_0 AVX optimization (ggerganov#9422)

  Show description for 5c3d0f1
  netrunnereve

  authored
  5c3d0f1

• cmake : do not hide GGML options + rename option (ggerganov#9465)

  Show description for 19514d6
  ggerganov

  authored
  19514d6

• convert : identify missing model files (ggerganov#9397)

  compilade

  authored
  d54c21d

• ggml : link MATH_LIBRARY not by its full path (ggerganov#9339)

  Xarbirus

  authored
  a6a3a5c

• llama : rename n_embed to n_embd in rwkv6_time_mix (ggerganov#9504)

  Show description for acb2c32
  danbev

  authored
  acb2c32

• ggml : move common CPU backend impl to new header (ggerganov#9509)

  slaren

  authored
  23e0d70

Commits on Sep 17, 2024

• llama : add llama_n_head() (ggerganov#9512)

  Xarbirus

  authored
  37f3a38

• llama : support IBM Granite architecture (ggerganov#9412)

  Show description for 0d2ec43
  gabe-l-hart

  authored
  0d2ec43

• unicode : add <algorithm> (ggerganov#9508)

  ykhrustalev

  authored
  503147a

• threadpool : skip polling for unused threads (ggerganov#9461)

  Show description for 0226613
  max-krasnyansky

  authored
  0226613

• llama : fix n_vocab init for 'no_vocab' case (ggerganov#9511)

  Show description for 8344ef5
  

  Xarbirus

  and

  ggerganov

  authored
  8344ef5

• arg : add env variable for parallel (ggerganov#9513)

  Show description for 8b836ae
  bertwagner

  authored
  8b836ae

• llama-bench: correct argument parsing error message (ggerganov#9524)

  Xarbirus

  authored
  7be099f

Commits on Sep 18, 2024

• [SYCL]set context default value to avoid memory issue, update guide (ggerganov#9476)

  Show description for faf67b3
  

  3 peopleauthored
  faf67b3

• server : fix OpenSSL build (remove obsolete LOG_INFO) (ggerganov#9529)

  EZForever

  authored
  f799155

• server : match OAI structured output response (ggerganov#9527)

  VJHack

  authored
  8a30835

• llama : use reserve/emplace_back in sampler_sample (ggerganov#9534)

  Show description for 6443ddd
  danbev

  authored
  6443ddd

• scripts : verify py deps at the start of compare (ggerganov#9520)

  ggerganov

  authored
  0d2f22e

• ggml : fix n_threads_cur initialization with one thread (ggerganov#9538)

  Show description for 64c6af3
  

  slaren

  and

  max-krasnyansky

  authored
  64c6af3

Commits on Sep 19, 2024

• imatrix : disable prompt escape by default (ggerganov#9543)

  CISC

  authored
  eca0fab

• server : clean-up completed tasks from waiting list (ggerganov#9531)

  Show description for 6026da5
  ggerganov

  authored
  6026da5

Commits on Sep 20, 2024

• perplexity : do not escape input data by default (ggerganov#9548)

  CISC

  authored
  722ec1e

• examples : flush log upon ctrl+c (ggerganov#9559)

  ggerganov

  authored
  d39e267

• CUDA: fix sum.cu compilation for CUDA < 11.7 (ggerganov#9562)

  JohannesGaessler

  authored
  5cb12f6

• examples : add null threadpool args where needed (ggml/0)

  Show description for a6809c6
  ggerganov

  committed
  a6809c6

• ggml/examples: add backend support for numerical optimization (ggml/949)

  Show description for 424c5d0
  

  3 peoplecommitted
  424c5d0

• sync : ggml

  Show description for 4301535
  ggerganov

  committed
  4301535

• ggml : fix trailing whitespace (#0)

  Show description for 27609c4
  ggerganov

  committed
  27609c4

• ggml : fix builds (#0)

  Show description for d13edb1
  ggerganov

  committed
  d13edb1

• quantize : improve type name parsing (ggerganov#9570)

  Show description for 6335114
  slaren

  authored
  6335114

Commits on Sep 21, 2024

• CI: Provide prebuilt windows binary for hip (ggerganov#9467)

  no1wudi

  authored
  e948a7d

• Update CUDA graph on scale change plus clear nodes/params (ggerganov#9550)

  Show description for 41f4778
  agray3

  authored
  41f4778

• ggml-alloc : fix list of allocated tensors with GGML_ALLOCATOR_DEBUG (ggerganov#9573)

  slaren

  authored
  d09770c

Commits on Sep 22, 2024

• RWKV v6: RWKV_WKV op CUDA implementation (ggerganov#9454)

  Show description for 2a63caa
  MollySophia

  authored
  2a63caa

• llama: remove redundant loop when constructing ubatch (ggerganov#9574)

  shankarg87

  authored
  ecd5d6b

• CUDA: enable Gemma FA for HIP/Pascal (ggerganov#9581)

  JohannesGaessler

  authored
  a5b57b0

• Fix merge error in ggerganov#9454 (ggerganov#9589)

  Show description for 912c331
  MollySophia

  authored
  912c331

• musa: enable building fat binaries, enable unified memory, and disable Flash Attention on QY1 (MTT S80) (ggerganov#9526)

  Show description for c35e586
  yeahdongcn

  authored
  c35e586

Commits on Sep 23, 2024

• Revert "[SYCL] fallback mmvq (ggerganov#9088)" (ggerganov#9579)

  Show description for e62e978
  

  Akarshan Biswas

  authored
  e62e978

• Merge branch 'layla-build' into merge

  l3utterfly

  authored
  ff3c559