ggerganov/whisper.cpp

Pull Requests Commits

examples : change exit code for requires_value_error function (#4105)

baiqieji123 committed 1 day ago

Verified 0c7be696

examples : fixes example #4106 issue (#4107)

Dj-pages committed 2 days ago

Verified 7e26d3b0

whisper : bump version to 1.9.5 (#4102)

danbev committed 7 days ago

Verified d1be6fde

talk-llama : update llama.cpp to v0.6.0

ggerganov committed 7 days ago

Verified 4afec37b

ggerganov committed 7 days ago

Verified 3d451abe

ggml : bump version to 0.26.0 (ggml/1652)

ggerganov committed 7 days ago

Verified 47f1c7dc

CUDA: make the alloc_deps check batch independent (llama/29986)

am17an committed 7 days ago

Verified e71b7844

vulkan: fix Flash Attention shmem write out of bounds (llama/29988)

0cc4m committed 7 days ago

Verified cdc87302

vulkan: revert mul_mat_id tile selection PR #29182 (llama/29936)

virajwad committed 7 days ago

Verified 563a9f7a

cuda: use the vector lightning indexer kernel on MUSA (llama/29990)

ServeurpersoCom committed 7 days ago

Verified ebb8a170

CUDA: Optimize accumulation in mmq for NVFP4 type (llama/29857)

kmorennv committed 7 days ago

Verified 8c6ca4ab

vulkan: fix stale prealloc_y reuse across flash attention and soft_max (llama/29591)

fxgsell committed 7 days ago

Verified 82fa9df7

vulkan: sparse flash attention for quantized K/V (llama/29639)

fxgsell committed 7 days ago

Verified f6915422

llama : fix unexpected graph reallocation in the k-pool models (llama/29958)

ggerganov committed 7 days ago

Verified d477de3c

ggml-cpu : add Q8_0 IME1 matrix kernel for SpacemiT X60 (llama/28479)

alanhc committed 7 days ago

Verified 95283ca7

vulkan : Fix undeclared identifiers when -DGGML_VULKAN_RUN_TESTS=ON (llama/29912)

EAddario committed 7 days ago

Verified 155c5c28

webgpu: add MMVQ support for Q1_0/Q5_0/Q5_1/Q3_K/Q5_K/Q6_K/MXFP4 (llama/29483)

yomaytk committed 7 days ago

Verified 7091d4a2

cuda: tile the lightning indexer over keys and tokens for 4 heads (llama/29901)

ServeurpersoCom committed 7 days ago

Verified 4cc62b81

metal : few-row MMA mat-mul (llama/29869)

pratiknarola-t committed 7 days ago

Verified 7e576cbd

CUDA: use MMVF for thin f16/bf16 mul_mat at small batch size (llama/29633)

ynankani committed 7 days ago

Verified 91da4702

CUDA: prefer whole-tile FlashAttention scheduling for efficient two-stage kernels (llama/29435)

anujj committed 7 days ago

Verified 8bb85d96

CUDA: refactor swizzling code (llama/29612)

JohannesGaessler committed 7 days ago

Verified ce6552ef

ggml-cpu: support BF16/FP16/FP32 K tails in tinyBLAS on x86 (llama/29806)

SongXiaoXi committed 7 days ago

Verified bfe3dc63

cuda : move neu_padded to where it is used (llama/29940)

angt committed 7 days ago

Verified 25453c0a

cuda : move blocks_per_col to where it is used (llama/29939)

angt committed 7 days ago

Verified 7914776a

CUDA: fix MMQ memory fault if n_expert >> n_ubatch (llama/29941)

JohannesGaessler committed 7 days ago

Verified 6421d2c1

vulkan: fix rdna4 mat_vec tuning (llama/29934)

0cc4m committed 7 days ago

Verified 05d4d89a

webgpu: add f16 support to fill/set_rows (llama/29897)

yomaytk committed 7 days ago

Verified 51bfbd61

ggml-openvino: update to 2026.4.1, optimize performance, expand ops, improve device listing. (llama/29852)

ravi9 committed 7 days ago

Verified f2aa80ad

qwen4exp : halve the indexer score memory (llama/29825)

ServeurpersoCom committed 7 days ago

Verified 7f4a67a5

Older