sync : ggml#3215

Merged

ggerganov merged 23 commits intomasterfrom

sync-ggml-25-06-01

Jun 1, 2025

Commits on Jun 1, 2025

ggml : remove ggml_graph_import and ggml_graph_export declarations (ggml/1247)

rgerganov
authored and
ggerganov
committed
cmake : Fix broken CMake error messages (ggml/1252)

dg0yt
authored and
ggerganov
committed
vulkan : Remove unexpected ; (ggml/1253)

dg0yt
authored and
ggerganov
committed
ggml : add ggml_repeat_4d (llama/13824)

ngxson
authored and
ggerganov
committed
SYCL: add gelu_erf kernel (llama/13749)

authored and
ggerganov
committed
vulkan: use timestamp queries for GGML_VULKAN_PERF (llama/13817)

jeffbolznv
authored and
ggerganov
committed
opencl: mark `mul_mat` `f32f32` as supporting non-contiguous tensors (llama/13790)

lhez
authored and
ggerganov
committed
opencl: add new ops - `argsort`, `div`, `sub`, `addrows`, `sigmoid`, `group_norm` (llama/13787)

lhez
authored and
ggerganov
committed
CANN: Add SOC TYPE printing in cmake configuration (llama/13837)

leo-pony
authored and
ggerganov
committed
CUDA: fix FA tg at long context for CC >= 8.9 (llama/13852)

JohannesGaessler
authored and
ggerganov
committed
ggml: aarch64: Implement SVE F32 kernels for vector functions (llama/13843)

vineelabhinav
authored and
ggerganov
committed
ggml: aarch64: Implement SVE F32 kernels for Mamba Sequential Scan Algorithm (llama/13882)

vineelabhinav
and
ggerganov
committed
cmake: Factor out CPU architecture detection (llama/13883)

ckastner
authored and
ggerganov
committed
arm64: optimize q4_k_q8_k kernel with i8mm (llama/13886)

cyb70289
authored and
ggerganov
committed
cmake: Guard GGML_CPU_ALL_VARIANTS by architecture (llama/13890)

ckastner
authored and
ggerganov
committed
SYCL: Add mrope kernel (llama/13755)

qnixsynapse
authored and
ggerganov
committed
cuda : prevent using split buffers with 3d/4d matrices (llama/13919)

slaren
authored and
ggerganov
committed
sched : avoid changing cur_copy when a graph is already allocated (llama/13922)

slaren
authored and
ggerganov
committed
CUDA: fix typo in FlashAttention code (llama/13926)

JohannesGaessler
authored and
ggerganov
committed
CUDA: add a prop in ggml_cuda_device_infor for distinguish iGPU or dGPU in cuda (#13856) (llama/13895)

authored and
ggerganov
committed
threading: support for GGML_SCHED_PRIO_LOW, update thread info on Windows to avoid throttling (llama/12995)

authored and
ggerganov
committed
sync : ggml
ggerganov
committed
talk-llama : sync llama.cpp
ggerganov
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

sync : ggml#3215

sync : ggml#3215
ggerganov merged 23 commits intomasterfrom
sync-ggml-25-06-01

Commits on Jun 1, 2025

ggml : remove ggml_graph_import and ggml_graph_export declarations (ggml/1247)

cmake : Fix broken CMake error messages (ggml/1252)

vulkan : Remove unexpected ; (ggml/1253)

ggml : add ggml_repeat_4d (llama/13824)

SYCL: add gelu_erf kernel (llama/13749)

vulkan: use timestamp queries for GGML_VULKAN_PERF (llama/13817)

opencl: mark `mul_mat` `f32f32` as supporting non-contiguous tensors (llama/13790)

opencl: add new ops - `argsort`, `div`, `sub`, `addrows`, `sigmoid`, `group_norm` (llama/13787)

CANN: Add SOC TYPE printing in cmake configuration (llama/13837)

CUDA: fix FA tg at long context for CC >= 8.9 (llama/13852)

ggml: aarch64: Implement SVE F32 kernels for vector functions (llama/13843)

ggml: aarch64: Implement SVE F32 kernels for Mamba Sequential Scan Algorithm (llama/13882)

cmake: Factor out CPU architecture detection (llama/13883)

arm64: optimize q4_k_q8_k kernel with i8mm (llama/13886)

cmake: Guard GGML_CPU_ALL_VARIANTS by architecture (llama/13890)

SYCL: Add mrope kernel (llama/13755)

cuda : prevent using split buffers with 3d/4d matrices (llama/13919)

sched : avoid changing cur_copy when a graph is already allocated (llama/13922)

CUDA: fix typo in FlashAttention code (llama/13926)

CUDA: add a prop in ggml_cuda_device_infor for distinguish iGPU or dGPU in cuda (#13856) (llama/13895)

threading: support for GGML_SCHED_PRIO_LOW, update thread info on Windows to avoid throttling (llama/12995)

sync : ggml

talk-llama : sync llama.cpp

sync : ggml#3215ggerganov merged 23 commits intomasterggml-org/whisper.cpp:masterfrom sync-ggml-25-06-01ggml-org/whisper.cpp:sync-ggml-25-06-01Copy head branch name to clipboard

Commits

Commits on Jun 1, 2025

sync : ggml#3215
ggerganov merged 23 commits intomasterfrom
sync-ggml-25-06-01