sync : ggml#3171

Merged

ggerganov merged 24 commits intomasterfrom

sync-whisper.cpp-25-05-19

May 19, 2025

Commits on May 19, 2025

ggml : Fix missing backtrace on Linux (ggml/1228)

danielzgtg
authored and
ggerganov
committed
ggml : fix apple OS check in ggml_print_backtrace (ggml/1229)

slaren
authored and
ggerganov
committed
mnist: fix segmentation fault (ggml/1227)

JohannesGaessler
authored and
ggerganov
committed
ggml-cpu: Update KleidiAI to v1.6 and fix include directives (llama/13509)

eddnjjn
authored and
ggerganov
committed
metal : optimize multi-sequence FA vec kernel (llama/13493)
ggerganov
committed
metal : use FA-vec kernel up to batch size 20 (llama/13496)
ggerganov
committed
vulkan: workaround FA compile failures on macos (llama/13517)

jeffbolznv
authored and
ggerganov
committed
vulkan: KHR_coopmat flash attention (llama/13506)

jeffbolznv
authored and
ggerganov
committed
cmake: simplify vulkan shader test logic (llama/13263)

bandoti
authored and
ggerganov
committed
CUDA: faster Deepseek FA, add Turing support (llama/13435)

JohannesGaessler
authored and
ggerganov
committed
CUDA: fix crash on large batch size for quant. MoE (llama/13537)

JohannesGaessler
authored and
ggerganov
committed
arm64: optimize q6_k_q8_k kernel with i8mm (llama/13519)

cyb70289
authored and
ggerganov
committed
sycl: use oneDNN for matrices multiplication (llama/12972)

lslusarczyk
authored and
ggerganov
committed
sycl: reordered Q4_K MMVQ (llama/13109)

sgeor255
authored and
ggerganov
committed
sycl: simplify bin_bcast_kernel (llama/13383)

AD2605
authored and
ggerganov
committed
gguf : use ggml log system (llama/13571)

slaren
authored and
ggerganov
committed
sycl : fixed compilation warnings (llama/13582)

lslusarczyk
authored and
ggerganov
committed
metal : add FA-vec kernel for head size 64 (llama/13583)
ggerganov
committed
vulkan: use scalar FA rather than coopmat2 when N==1 (llama/13554)

jeffbolznv
authored and
ggerganov
committed
vulkan: move common FA code to flash_attn_base.comp (llama/13556)

jeffbolznv
authored and
ggerganov
committed
cmake: use the current build config for vulkan-shaders-gen (llama/13595)

giladgd
authored and
ggerganov
committed
CANN: Support MOE Model MUL_MAT_ID (llama/13042)

noemotiovon
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#3171

sync : ggml#3171
ggerganov merged 24 commits intomasterfrom
sync-whisper.cpp-25-05-19

Commits on May 19, 2025

ggml : Fix missing backtrace on Linux (ggml/1228)

ggml : fix apple OS check in ggml_print_backtrace (ggml/1229)

mnist: fix segmentation fault (ggml/1227)

ggml-cpu: Update KleidiAI to v1.6 and fix include directives (llama/13509)

metal : optimize multi-sequence FA vec kernel (llama/13493)

metal : use FA-vec kernel up to batch size 20 (llama/13496)

vulkan: workaround FA compile failures on macos (llama/13517)

vulkan: KHR_coopmat flash attention (llama/13506)

cmake: simplify vulkan shader test logic (llama/13263)

CUDA: faster Deepseek FA, add Turing support (llama/13435)

CUDA: fix crash on large batch size for quant. MoE (llama/13537)

arm64: optimize q6_k_q8_k kernel with i8mm (llama/13519)

sycl: use oneDNN for matrices multiplication (llama/12972)

sycl: reordered Q4_K MMVQ (llama/13109)

sycl: simplify bin_bcast_kernel (llama/13383)

gguf : use ggml log system (llama/13571)

sycl : fixed compilation warnings (llama/13582)

metal : add FA-vec kernel for head size 64 (llama/13583)

vulkan: use scalar FA rather than coopmat2 when N==1 (llama/13554)

vulkan: move common FA code to flash_attn_base.comp (llama/13556)

cmake: use the current build config for vulkan-shaders-gen (llama/13595)

CANN: Support MOE Model MUL_MAT_ID (llama/13042)

sync : ggml

talk-llama : sync llama.cpp

sync : ggml#3171ggerganov merged 24 commits intomasterggml-org/whisper.cpp:masterfrom sync-whisper.cpp-25-05-19ggml-org/whisper.cpp:sync-whisper.cpp-25-05-19Copy head branch name to clipboard

Commits

Commits on May 19, 2025

sync : ggml#3171
ggerganov merged 24 commits intomasterfrom
sync-whisper.cpp-25-05-19