Commits

Branch selector

 7bed7eb

User selector

All users

Datepicker

All time

Commit History

Commits on Jan 3, 2024

• cuda : simplify expression

  Show description for 7bed7eb
  

  ggerganov

  and

  slaren

  committed
  7bed7eb

  Copy full SHA for 7bed7eb

• cuda : mark I16 and I32 ops as unsupported

  Show description for d55356d
  ggerganov

  committed
  d55356d

  Copy full SHA for d55356d

• sync : ggml

  Show description for 75e3fd8
  ggerganov

  committed
  75e3fd8

  Copy full SHA for 75e3fd8

• metal : add kernel_get_rows_i32

  Show description for 2893137
  ggerganov

  committed
  2893137

  Copy full SHA for 2893137

• scripts : fix sync order + metal sed

  ggerganov

  committed
  ab62fc3

  Copy full SHA for ab62fc3

• ggml : extend ggml_get_rows, ggml_repeat, ggml_concat (ggml/639)

  Show description for 5f66ebc
  

  gwenzek

  and

  ggerganov

  committed
  5f66ebc

  Copy full SHA for 5f66ebc

• server : throw an error when slot unavailable (ggerganov#4741)

  jparkerweb

  authored
  f2eb19b

  Copy full SHA for f2eb19b

Commits on Jan 2, 2024

• metal : optimize ggml_mul_mat_id (faster Mixtral PP) (ggerganov#4725)

  Show description for f3f62f0
  ggerganov

  authored
  f3f62f0

  Copy full SHA for f3f62f0

• server : add token counts to html footer (ggerganov#4738)

  Show description for 0ef3ca2
  phiharri

  authored
  0ef3ca2

  Copy full SHA for 0ef3ca2

• llama : llama_model_desc print number of experts

  ggerganov

  committed
  540938f

  Copy full SHA for 540938f

• llama : replace all API facing int's with int32_t (ggerganov#4577)

  Show description for 0040d42
  MarcusDunn

  authored
  0040d42

  Copy full SHA for 0040d42

• llama : differentiate the KV dims in the attention (ggerganov#4657)

  Show description for 83e633c
  

  postmasters

  and

  ggerganov

  authored
  83e633c

  Copy full SHA for 83e633c

• editorconfig : fix whitespace and indentation ggerganov#4710

  ggerganov

  committed
  32866c5

  Copy full SHA for 32866c5

• server : add --override-kv parameter (ggerganov#4710)

  Show description for 5d7002d
  

  3 peopleauthored
  5d7002d

  Copy full SHA for 5d7002d

• py : re-enable mmap in convert hf (ggerganov#4732)

  Show description for 26f3071
  

  4 peopleauthored
  26f3071

  Copy full SHA for 26f3071

• finetune: fix typo in README.md (ggerganov#4733)

  Show description for 775ac87
  danbev

  authored
  775ac87

  Copy full SHA for 775ac87

• metal : enable shader debugging (cmake option) (ggerganov#4705)

  Show description for 58ba655
  ggerganov

  authored
  58ba655

  Copy full SHA for 58ba655

Commits on Dec 31, 2023

• flake.lock: update

  Show description for edd1ab7
  

  SomeoneSerge

  authored and

  philiptaron

  committed
  edd1ab7

  Copy full SHA for edd1ab7

• flake.nix: suggest the binary caches

  

  SomeoneSerge

  authored and

  philiptaron

  committed
  198ed7e

  Copy full SHA for 198ed7e

• workflows: nix-ci: add a qemu job for jetsons

  

  SomeoneSerge

  authored and

  philiptaron

  committed
  d836174

  Copy full SHA for d836174

• workflows: nix-flakestry: drop tag filters

  Show description for 06f2a5d
  

  SomeoneSerge

  authored and

  philiptaron

  committed
  06f2a5d

  Copy full SHA for 06f2a5d

• workflows: weekly nix flake update

  

  SomeoneSerge

  authored and

  philiptaron

  committed
  c523994

  Copy full SHA for c523994

• workflows: nix-ci: add a job for eval

  

  SomeoneSerge

  authored and

  philiptaron

  committed
  1e9ae54

  Copy full SHA for 1e9ae54

• workflows: nix-ci: init; build flake outputs

  

  SomeoneSerge

  authored and

  philiptaron

  committed
  7adedec

  Copy full SHA for 7adedec

• flake.nix: expose checks

  

  SomeoneSerge

  authored and

  philiptaron

  committed
  356ea17

  Copy full SHA for 356ea17

• flake.nix: rocm not yet supported on aarch64, so hide the output

  

  SomeoneSerge

  authored and

  philiptaron

  committed
  a5c088d

  Copy full SHA for a5c088d

• flake.nix: expose full scope in legacyPackages

  

  SomeoneSerge

  authored and

  philiptaron

  committed
  1e3900e

  Copy full SHA for 1e3900e

• ggml : add ggml_vdotq_s32 alias (ggerganov#4715)

  Show description for e39106c
  ggerganov

  authored
  e39106c

  Copy full SHA for e39106c

Commits on Dec 30, 2023

• clip : refactor + bug fixes (ggerganov#4696)

  Show description for 9fbda71
  ggerganov

  authored
  9fbda71

  Copy full SHA for 9fbda71

• CUDA: fixed tensor cores not being used on RDNA3 (ggerganov#4697)

  JohannesGaessler

  authored
  39d8bc7

  Copy full SHA for 39d8bc7

• ggml : add ggml_cpu_has_avx_vnni() (ggerganov#4589)

  Show description for 24a447e
  

  tikikun

  and

  ggerganov

  authored
  24a447e

  Copy full SHA for 24a447e

Commits on Dec 29, 2023

• CUDA: fix tensor core logic for Pascal and HIP (ggerganov#4682)

  JohannesGaessler

  authored
  a20f3c7

  Copy full SHA for a20f3c7

• clip : use ggml_backend_buffer_is_host (ggerganov#4205)

  ggerganov

  committed
  0235b9b

  Copy full SHA for 0235b9b

• clip : enable gpu backend (ggerganov#4205)

  Show description for ce18d72
  

  FSSRepo

  and

  ggerganov

  authored
  ce18d72

  Copy full SHA for ce18d72

• cuda: fix vmm oom issue on NVIDIA AGX Orin (ggerganov#4687)

  Show description for 91bb39c
  hydai

  authored
  91bb39c

  Copy full SHA for 91bb39c

Pagination

Previous

Next