ggml-org/whisper.cpp

Pull Requests Commits

vulkan : reuse descriptor sets when bindings are constant (llama/29280)

realrasengan committed 9 days ago

Verified eecee67b

vulkan: include functional header (llama/29597)

LinuxUserGD committed 9 days ago

Verified 97690105

ggml-openvino: mark unaligned batch-stride views unsupported (llama/29603)

ravi9 committed 9 days ago

Verified 830dc664

webgpu: Handle unaligned writes in ggml_backend_webgpu_buffer_set_tensor (llama/29471)

jbooth committed 9 days ago

Verified 81b951e0

tests : refactor test-recurrent-state-rollback (llama/29426)

ggerganov committed 9 days ago

Verified ee86a150

ggml-cpu: enable tiled flash attention for non-vector-multiple head dims on x86 (llama/29423)

SongXiaoXi committed 9 days ago

Verified 83fd1554

vulkan: fuse qwen4exp's SCALE -> SIGMOID -> SCALE -> hc_post chain (llama/29520)

fxgsell committed 9 days ago

Verified 65449dbc

HIP: fix template skip for DKQ > 256 mfma kernels (llama/29559)

IMbackK committed 9 days ago

Verified 0c17f7d6

metal: support left and circular padding in GGML_OP_PAD (llama/29561)

ServeurpersoCom committed 9 days ago

Verified 0e537510

Enables Windows ARM64 build with MSVC cl.exe (llama/28362)

sarahwu185 committed 9 days ago

Verified b7b84950

vulkan: fix wrong results when a mul_mat reads a slice of a larger cache (llama/28956)

ServeurpersoCom committed 9 days ago

Verified 85f69261

opencl: refine bin kernel loading condition (llama/29503)

lhez committed 9 days ago

Verified 51db4755

sycl: FWHT kernels for block widths above 512 (llama/29243)

bri-prism committed 9 days ago

Verified 55a695da

CUDA: tune fp16 tile FlashAttention configs for head sizes 40-112 (llama/26289)

animeshsri14 committed 9 days ago

Verified 846af526

HIP: Enable fattn-mma kernel on cdna for dkq > 256 for large batch sizes (llama/28907)

IMbackK committed 9 days ago

Verified 743f1ad2

vulkan: fix argsort kernel selection for Adreno (llama/29469)

0cc4m committed 9 days ago

Verified 14d1aa70

RPC: use RDMA completion channel to not spin (llama/29440)

am17an committed 9 days ago

Verified 24cf265b

hexagon: support tiled Q4_0 and Q8_0 GET_ROWS (llama/29511)

kurquhar committed 9 days ago

Verified 7eea5188

hexagon: support for backend sampler (llama/29502)

max-krasnyansky committed 9 days ago

Verified fe06027d

cuda: support Nemotron 3 Puzzle state size 96 for ssm scan (llama/28717)

anavp-nvidia committed 9 days ago

Verified fc1ebfab

cuda: add F16 input to the FWHT (llama/29096)

bri-prism committed 9 days ago

Verified dfe8fbd7

ggml-cpu: tiled mul_mat for k-quants (llama/27851)

jbooth committed 9 days ago

Verified 447a7751

opencl: add A8 Q8_0 non-MoE dp4a binary kernel (llama/29439)

shaofeiqi committed 9 days ago

Verified bf1d787a

hexagon: find software divide calls using binary inspection tool (llama/29449)

trivikram-reddy1 committed 9 days ago

Verified 9731f6fd

opencl: add bin kernel `kernel_gemm_noshuffle_q5_k_f32_32b_trans_ila_a8_bin`, `kernel_gemm_noshuffle_q5_k_q8_1_dp4a_ila_a8_bin` (llama/29401)

shaofeiqi committed 9 days ago

Verified 2d91497a

Fixing the vulkan build issue of legacy GLSLC version that has no cooperativeMatrix API support (https://github.com/ggml-org/llama.cpp/issues/29373) (llama/29409)

sliu39 committed 9 days ago

Verified 669188ef

metal: FWHT kernels for block widths above 512 (llama/29095)

bri-prism committed 9 days ago

Verified 9a919424

ggerganov committed 9 days ago

Verified 331d3c8a

metal : split fa kernels into per-dtype libraries (llama/29329)

ggerganov committed 9 days ago

Verified 4b0998ad

ggerganov committed 9 days ago

Verified bad9b546

Newer Older