sync : ggml#3329

Merged

ggerganov merged 20 commits intomasterfrom

sync-ggml-25-07-19

Jul 19, 2025

Commits on Jul 19, 2025

CUDA: add set rows for f32 and f16 (llama/14551)

am17an
authored and
ggerganov
committed
metal : Add missing unary ops Metal support (llama/14660)

YavorGIvanov
authored and
ggerganov
committed
ggml : add build-time message to remind about ggml_set_rows (llama/14661)
ggerganov
committed
cuda : add ELU support (llama/14657)

YavorGIvanov
authored and
ggerganov
committed
cuda : add set rows for bf16 (llama/14664)

CISC
authored and
ggerganov
committed
sycl: Batched mulmat rework for oneDNN dispatch (llama/14617)

ShanoToni
authored and
ggerganov
committed
SYCL: use 1D kernel for set_rows (llama/14618)

qnixsynapse
authored and
ggerganov
committed
ggml : refactor llamafile_sgemm PPC code (llama/14673)

shalinib-ibm
authored and
ggerganov
committed
sycl: Hotfix for non dnnl codepath (llama/14677)

ShanoToni
authored and
ggerganov
committed
cuda: fix build warnings in set-rows.cu (unused variable) (llama/14687)

yeahdongcn
authored and
ggerganov
committed
vulkan: add RTE variants for glu/add/sub/mul/div (llama/14653)

jeffbolznv
authored and
ggerganov
committed
vulkan: fix noncontig check for mat_mul_id splitting (llama/14683)

jeffbolznv
authored and
ggerganov
committed
ggml : add asserts (llama/14720)

ggerganov
and
slaren
committed
llama : add high-throughput mode (llama/14363)

ggerganov
and
JohannesGaessler
committed
ggml: Add initial WebGPU backend (llama/14521)

reeselevine
authored and
ggerganov
committed
use max work group size for device to replace the magic number (llama/14732)

NeoZhangJianyu
authored and
ggerganov
committed
CUDA: set_rows + cpy.cu refactor (llama/14712)

am17an
authored and
ggerganov
committed
cuda : Fix Gemma3n not executed as CUDA_GRAPH on NVGPUs (llama/14741)

ORippler
authored and
ggerganov
committed
metal : fuse add, mul + add tests (llama/14596)
ggerganov
committed
sync : ggml
ggerganov
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

sync : ggml#3329

sync : ggml#3329
ggerganov merged 20 commits intomasterfrom
sync-ggml-25-07-19

Commits on Jul 19, 2025

CUDA: add set rows for f32 and f16 (llama/14551)

metal : Add missing unary ops Metal support (llama/14660)

ggml : add build-time message to remind about ggml_set_rows (llama/14661)

cuda : add ELU support (llama/14657)

cuda : add set rows for bf16 (llama/14664)

sycl: Batched mulmat rework for oneDNN dispatch (llama/14617)

SYCL: use 1D kernel for set_rows (llama/14618)

ggml : refactor llamafile_sgemm PPC code (llama/14673)

sycl: Hotfix for non dnnl codepath (llama/14677)

cuda: fix build warnings in set-rows.cu (unused variable) (llama/14687)

vulkan: add RTE variants for glu/add/sub/mul/div (llama/14653)

vulkan: fix noncontig check for mat_mul_id splitting (llama/14683)

ggml : add asserts (llama/14720)

llama : add high-throughput mode (llama/14363)

ggml: Add initial WebGPU backend (llama/14521)

use max work group size for device to replace the magic number (llama/14732)

CUDA: set_rows + cpy.cu refactor (llama/14712)

cuda : Fix Gemma3n not executed as CUDA_GRAPH on NVGPUs (llama/14741)

metal : fuse add, mul + add tests (llama/14596)

sync : ggml

sync : ggml#3329ggerganov merged 20 commits intomasterggml-org/whisper.cpp:masterfrom sync-ggml-25-07-19ggml-org/whisper.cpp:sync-ggml-25-07-19Copy head branch name to clipboard

Commits

Commits on Jul 19, 2025

sync : ggml#3329
ggerganov merged 20 commits intomasterfrom
sync-ggml-25-07-19