Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Oct 22, 2025

• sync : whisper.cpp

  Show description for 2d3876d
  ggerganov

  committed
  2d3876d

  Copy full SHA for 2d3876d

Commits on Oct 21, 2025

• sync : llama.cpp

  ggerganov

  committed
  999574b

  Copy full SHA for 999574b

• ggml: add ggml_can_fuse_subgraph (llama/16662)

  Show description for b0b980a
  

  am17an

  authored and

  ggerganov

  committed
  b0b980a

  Copy full SHA for b0b980a

• opencl: fix warnings and clean up profiling (llama/16688)

  Show description for 2a2ffcf
  

  lhez

  authored and

  ggerganov

  committed
  2a2ffcf

  Copy full SHA for 2a2ffcf

• vulkan: Handle FA with all -inf mask values (llama/16447)

  

  jeffbolznv

  authored and

  ggerganov

  committed
  a118c51

  Copy full SHA for a118c51

• sycl : add PAD_REFLECT_D1 operator support (llama/16145)

  Show description for 53b34f8
  

  ye-NX

  authored and

  ggerganov

  committed
  53b34f8

  Copy full SHA for 53b34f8

• ggml-alloc : fix leak when reusing a tensor with a larger size (llama/16679)

  

  slaren

  authored and

  ggerganov

  committed
  9dea367

  Copy full SHA for 9dea367

• SYCL: Add support for FLOOR,CEIL,ROUND and TRUNC unary operators (llama/16613)

  Show description for 95e5e82
  

  safranowith

  authored and

  ggerganov

  committed
  95e5e82

  Copy full SHA for 95e5e82

• ci : fix binaries release failure for s390x (binaries may not work yet) (llama/16664)

  Show description for 2f2a9b9
  

  taronaeo

  authored and

  ggerganov

  committed
  2f2a9b9

  Copy full SHA for 2f2a9b9

• HIP: fix GPU_TARGETS (llama/16642)

  

  JohannesGaessler

  authored and

  ggerganov

  committed
  a3fde8d

  Copy full SHA for a3fde8d

• vulkan: Implement topk_moe fused shader, ported from CUDA (llama/16641)

  Show description for c48d0af
  

  jeffbolznv

  authored and

  ggerganov

  committed
  c48d0af

  Copy full SHA for c48d0af

• CUDA: use registers instead of smem in topk-moe (llama/16647)

  Show description for 1fe7675
  

  am17an

  authored and

  ggerganov

  committed
  1fe7675

  Copy full SHA for 1fe7675

• opencl: transposed gemm/gemv moe kernel with mxfp4,f32 (llama/16602)

  Show description for 90fac5e
  

  shawngu-quic

  authored and

  ggerganov

  committed
  90fac5e

  Copy full SHA for 90fac5e

• rpc : report actual free memory (llama/16616)

  Show description for 37ac04e
  

  rgerganov

  authored and

  ggerganov

  committed
  37ac04e

  Copy full SHA for 37ac04e

• vulkan: Add State Space Model (SSM) Operations Support (llama/16463)

  Show description for fae740a
  

  giuseppe

  authored and

  ggerganov

  committed
  fae740a

  Copy full SHA for fae740a

• ggml : fix SpaceMit IME array out-of-bounds in task assignment (llama/16629)

  Show description for 40f0011
  

  2 peopleauthored and

  ggerganov

  committed
  40f0011

  Copy full SHA for 40f0011

• vulkan: fix debug build (add_rms_len/data not found) (llama/16624)

  

  jeffbolznv

  authored and

  ggerganov

  committed
  8c30a31

  Copy full SHA for 8c30a31

• metal : add `CONV_TRANSPOSE_2D` (llama/16542)

  Show description for d86a5df
  

  iliailmer

  and

  ggerganov

  committed
  d86a5df

  Copy full SHA for d86a5df

• SYCL SET operator optimized for F32 tensors (llama/16350)

  Show description for b15a4a1
  

  2 peopleauthored and

  ggerganov

  committed
  b15a4a1

  Copy full SHA for b15a4a1

• sycl : add ARANGE operator (llama/16362)

  Show description for f8174e2
  

  2 peopleauthored and

  ggerganov

  committed
  f8174e2

  Copy full SHA for f8174e2

• CANN: format code using .clang-format (llama/15863)

  Show description for d573fb7
  

  2 peopleauthored and

  ggerganov

  committed
  d573fb7

  Copy full SHA for d573fb7

• ggml-cpu: replace putenv with setenv for const-correctness (llama/16573)

  Show description for a6ad99e
  

  otegami

  authored and

  ggerganov

  committed
  a6ad99e

  Copy full SHA for a6ad99e

• SYCL: Add GGML_OP_MEAN operator support (llama/16009)

  Show description for a85bfce
  

  2 peopleauthored and

  ggerganov

  committed
  a85bfce

  Copy full SHA for a85bfce

• cpu : add FLOOR, CEIL, ROUND and TRUNC unary operators (llama/16083)

  Show description for 9715e19
  

  safranowith

  authored and

  ggerganov

  committed
  9715e19

  Copy full SHA for 9715e19

• opencl: add q8_0 mm support (llama/16469)

  Show description for 4560236
  

  lhez

  authored and

  ggerganov

  committed
  4560236

  Copy full SHA for 4560236

• opencl: fix FA for f32 (llama/16584)

  

  lhez

  authored and

  ggerganov

  committed
  70e35ab

  Copy full SHA for 70e35ab

• metal: optimise `GGML_OP_SUM` (llama/16559)

  Show description for 270da01
  

  cern1710

  and

  ggerganov

  committed
  270da01

  Copy full SHA for 270da01

• CUDA: Changing the CUDA scheduling strategy to spin (llama/16585)

  Show description for ca02fcb
  

  2 peopleauthored and

  ggerganov

  committed
  ca02fcb

  Copy full SHA for ca02fcb

• metal : avoid using Metal's gpuAddress property (llama/16576)

  Show description for e416dab
  ggerganov

  committed
  e416dab

  Copy full SHA for e416dab

Commits on Oct 14, 2025

• sync : llama.cpp

  ggerganov

  committed
  c538174

  Copy full SHA for c538174

• vulkan: Add ACC_TYPE_VEC2 implementation (llama/16203)

  Show description for 98a8312
  

  2 peopleauthored and

  ggerganov

  committed
  98a8312

  Copy full SHA for 98a8312

• CUDA + openCL: fix bug in accessing rms_norm->src while doing fusion (llama/16577)

  

  am17an

  authored and

  ggerganov

  committed
  bbf742e

  Copy full SHA for bbf742e

• vulkan: Support FA with K/V in F32 (llama/16543)

  

  jeffbolznv

  authored and

  ggerganov

  committed
  73ccb8a

  Copy full SHA for 73ccb8a

• vulkan: Improve build time for MSVC (llama/16545)

  Show description for f110aec
  

  jeffbolznv

  authored and

  ggerganov

  committed
  f110aec

  Copy full SHA for f110aec

• CUDA: enable FA for FP32 KV cache (llama/16546)

  

  JohannesGaessler

  authored and

  ggerganov

  committed
  a499133

  Copy full SHA for a499133

Pagination

Previous

Next