ggml-org/whisper.cpp

Pull Requests Commits

llama : add `llama_prec_policy` + model-driven W4A4 path (llama/24364)

ynankani committed 12 days ago

Verified afc10c26

HIP: bump HIP_VERSION requried for fp8 to avoid missing __hip_fp8_e4m3 support in 6.2 (llama/29231)

IMbackK committed 12 days ago

Verified d7e83b18

rpc: include nb in the get_alloc_size cache key and floor the result at ggml_nbytes (llama/29283)

Jesssullivan committed 12 days ago

Verified 26bfdc66

support sparse FA (llama/28796)

arthw committed 12 days ago

Verified 93c4308a

musa: fix PH1 (MTT S5000) operator failures and build issues (llama/29193)

yeahdongcn committed 12 days ago

Verified f7d5bf38

CUDA: fuse RMS_NORM + SCALE into one kernel (llama/29393)

InflexCZE committed 12 days ago

Verified 40c872b2

hexagon: add q5_k quant type support (llama/29123)

jhen0409 committed 12 days ago

Verified 0458c9eb

hexagon: use DMA for contiguous dim1 CONCAT (llama/29404)

kurquhar committed 12 days ago

Verified eea9aa63

metal : fix graph capture and handle empty graphs (llama/29390)

ggerganov committed 12 days ago

Verified 4550b9db

metal : optimize sparse FA + clean-up (llama/29377)

ggerganov committed 12 days ago

Verified e1b93a51

hexagon: handle multi-sequence in concat_2d (llama/29344)

jhen0409 committed 12 days ago

Verified 61032a2b

hexagon: dynamic quantizer improvements (llama/29395)

kurquhar committed 12 days ago

Verified 845d1c02

hexagon: support I32 CPY and CONT (llama/29379)

kurquhar committed 12 days ago

Verified e5a88776

cuda : add F16 kernel support for CONV_2D_DW (llama/29064)

crion99 committed 12 days ago

Verified 5a9a3c06

vulkan: int8 coopmat1 matmul implementation for AMD RDNA3 and RDNA4 (llama/27952)

0cc4m committed 12 days ago

Verified cf0852d2

ggml : bump version to 0.25.3 (ggml/1645)

ggerganov committed 12 days ago

Verified 6d0197cc

ggml : fix ubsan error in `ggml_graph_nbytes` (ggml/1644)

ggerganov committed 12 days ago

Verified a2b9a22e

ggml : bump version to 0.25.2 (ggml/1642)

ggerganov committed 12 days ago

Verified 7ab1529b

vulkan: handle misalignment in conv_2d and conv_3d (llama/29365)

0cc4m committed 12 days ago

Verified f938f319

vulkan: tune KHR cooperative matrix support for Adreno GPUs (llama/29328)

Raman-Raje committed 12 days ago

Verified bf8128ec

cuda : add conv3d with implicit GEMM (llama/29137)

leejet committed 12 days ago

Verified 3ef6a832

hexagon: reject MUL_MAT_ID when src1 precision is F32 (llama/29348)

jhen0409 committed 12 days ago

Verified 6f2a84af

opencl: add A8 Q6_K non-MoE dp4a binary kernel (llama/29057)

shaofeiqi committed 12 days ago

Verified 0adffd58

whisper : add check for ttype in whisper and parakeet (#4091)

renvins committed 16 days ago

Verified 60c0be6a

cli : fix incorrect error code when files are failing (#4080)

renvins committed 20 days ago

Verified 6e4ab854

ci : cover GGML_BACKEND_DL in ubuntu-22-clang-arm64 (#4047)

apollo-2006 committed 24 days ago

Verified d09f61a7

ggerganov committed 25 days ago

a664346e

ggml : bump version to 0.25.1 (ggml/1637)

ggerganov committed 25 days ago

84f080e9

CUDA: add a reserve to avoid spurious warning on older GCC builds (llama/29317)

am17an committed 25 days ago

8f5ac2a6

metal: add the missing f32 x bf16 mul_mv variants (llama/28741)

ServeurpersoCom committed 25 days ago

8917ea04

Newer Older