Commits

Breadcrumbs

History for

onc6c4fc0

User selector

All users

Datepicker

All time

Commit History

Commits on Dec 14, 2023

• ggml : use ggml_row_size where possible (ggerganov#4472)

  Show description for 6744dbe
  slaren

  authored
  6744dbe

  Copy full SHA for 6744dbe

Commits on Dec 13, 2023

• sync : ggml (SD ops, tests, kernels) (ggerganov#4444)

  Show description for 4d98d9a
  

  ggerganov

  and

  slaren

  authored
  4d98d9a

  Copy full SHA for 4d98d9a

• llama : add Mixtral support (ggerganov#4406)

  Show description for 799a1cb
  

  3 peopleauthored
  799a1cb

  Copy full SHA for 799a1cb

Commits on Dec 7, 2023

• sync : ggml (new ops, tests, backend, etc.) (ggerganov#4359)

  Show description for fe680e3
  

  ggerganov

  and

  slaren

  authored
  fe680e3

  Copy full SHA for fe680e3

• llama : per-layer KV cache + quantum K cache (ggerganov#4309)

  Show description for bcc0eb4
  

  ggerganov

  and

  slaren

  authored
  bcc0eb4

  Copy full SHA for bcc0eb4

Commits on Dec 1, 2023

• ggml : add ggml_soft_max_ext (ggerganov#4256)

  Show description for ef47ec1
  ggerganov

  authored
  ef47ec1

  Copy full SHA for ef47ec1

Commits on Nov 24, 2023

• ggml-cuda : support stablelm rope (ggerganov#4156)

  Show description for 8a052c1
  

  slaren

  and

  ggerganov

  authored
  8a052c1

  Copy full SHA for 8a052c1

Commits on Nov 23, 2023

• Fix incorrect format strings and uninitialized variables. (ggerganov#4133)

  Show description for 55978ce
  haohui

  authored
  55978ce

  Copy full SHA for 55978ce

Commits on Nov 18, 2023

• Clean up ggml-cuda.cu warnings when compiling with clang (for ROCM) (ggerganov#4124)

  Show description for 2923f17
  KerfuffleV2

  authored
  2923f17

  Copy full SHA for 2923f17

Commits on Nov 17, 2023

• cuda : get_row_rounding F32 (ggerganov#4095)

  Show description for b83e149
  

  AndrewGodfrey

  and

  cebtenzzre

  authored
  b83e149

  Copy full SHA for b83e149

• llama : fix data units (ggerganov#4101)

  Show description for 4f447a4
  ggerganov

  authored
  4f447a4

  Copy full SHA for 4f447a4

Commits on Nov 15, 2023

• ggml-cuda : increase max graph size (ggerganov#4084)

  slaren

  authored
  1cf2850

  Copy full SHA for 1cf2850

Commits on Nov 13, 2023

• ggml : sync (im2col, GPU conv, 32-bit arm compat) (ggerganov#4060)

  Show description for 3d68f36
  ggerganov

  authored
  3d68f36

  Copy full SHA for 3d68f36

• sync : ggml (backend v2) (ggerganov#3912)

  Show description for 4760e7c
  ggerganov

  authored
  4760e7c

  Copy full SHA for 4760e7c

• Add ReLU and SQR CUDA ops to (partially) fix Persimmon offloading (ggerganov#4041)

  Show description for bb50a79
  KerfuffleV2

  authored
  bb50a79

  Copy full SHA for bb50a79

Commits on Nov 7, 2023

• cuda : supports running on CPU for GGML_USE_CUBLAS=ON build (ggerganov#3946)

  Show description for 46876d2
  wsxiaoys

  authored
  46876d2

  Copy full SHA for 46876d2

Commits on Nov 5, 2023

• ggml-cuda : fix f16 mul mat (ggerganov#3961)

  Show description for 2833a6f
  slaren

  authored
  2833a6f

  Copy full SHA for 2833a6f

• cuda : fix disabling device with --tensor-split 1,0 (ggerganov#3951)

  Show description for 132d25b
  

  cebtenzzre

  and

  slaren

  authored
  132d25b

  Copy full SHA for 132d25b

• cuda : revert CUDA pool stuff (ggerganov#3944)

  Show description for 48ade94
  slaren

  authored
  48ade94

  Copy full SHA for 48ade94

Commits on Nov 3, 2023

• ggml-cuda : move row numbers to x grid dim in mmv kernels (ggerganov#3921)

  slaren

  authored
  abb77e7

  Copy full SHA for abb77e7

Commits on Nov 2, 2023

• cuda : add ROCM aliases for CUDA pool stuff (ggerganov#3918)

  KerfuffleV2

  authored
  629f917

  Copy full SHA for 629f917

• cuda : fix const ptrs warning causing ROCm build issues (ggerganov#3913)

  ggerganov

  authored
  c7743fe

  Copy full SHA for c7743fe

• cuda : use CUDA memory pool with async memory allocation/deallocation when available (ggerganov#3903)

  Show description for d606905
  

  young-developer

  and

  Oleksii Maryshchenko

  authored
  d606905

  Copy full SHA for d606905

• cuda : check if this fixes Pascal card regression (ggerganov#3882)

  ggerganov

  authored
  4d719a6

  Copy full SHA for 4d719a6

• cuda : fix RoPE after ggerganov#2268 (ggerganov#3897)

  cebtenzzre

  authored
  2fffa0d

  Copy full SHA for 2fffa0d

Commits on Nov 1, 2023

• ggml-cuda : compute ptrs for cublasGemmBatchedEx in a kernel (ggerganov#3891)

  Show description for d02e98c
  slaren

  authored
  d02e98c

  Copy full SHA for d02e98c

• llama : implement YaRN RoPE scaling (ggerganov#2268)

  Show description for 898aeca
  

  cebtenzzre

  and

  jquesnelle

  authored
  898aeca

  Copy full SHA for 898aeca

• finetune : add -ngl parameter (ggerganov#3762)

  Show description for 73bdcb3
  

  3 peopleauthored
  73bdcb3

  Copy full SHA for 73bdcb3

Commits on Oct 27, 2023

• cuda : improve text-generation and batched decoding performance (ggerganov#3776)

  Show description for 2f9ec7e
  ggerganov

  authored
  2f9ec7e

  Copy full SHA for 2f9ec7e

Commits on Oct 25, 2023

• batched-bench : print params at start

  ggerganov

  committed
  6961c4b

  Copy full SHA for 6961c4b

Commits on Oct 24, 2023

• sync : ggml (conv ops + cuda MSVC fixes) (ggerganov#3765)

  Show description for b2f7e04
  ggerganov

  authored
  b2f7e04

  Copy full SHA for b2f7e04

• cuda : add batched cuBLAS GEMM for faster attention (ggerganov#3749)

  Show description for 2b4ea35
  

  ggerganov

  and

  KerfuffleV2

  authored
  2b4ea35

  Copy full SHA for 2b4ea35

Commits on Oct 10, 2023

• llm : add MPT support (ggerganov#3417)

  Show description for f5f9121
  

  3 peopleauthored
  f5f9121

  Copy full SHA for f5f9121

Commits on Oct 8, 2023

• sync : ggml (ggml-backend) (ggerganov#3548)

  Show description for db3abcc
  ggerganov

  authored
  db3abcc

  Copy full SHA for db3abcc

Commits on Sep 30, 2023

• ggml-cuda : perform cublas mat mul of quantized types as f16 (ggerganov#3412)

  Show description for f5ef5cf
  slaren

  authored
  f5ef5cf

  Copy full SHA for f5ef5cf

Pagination

Previous

Next