ggml-org/whisper.cpp

Pull Requests Commits

talk-llama : sync llama.cpp

ggerganov committed 2 days ago

0ae02cdb

ggerganov committed 2 days ago

0279b538

ggml : bump version to 0.15.3 (ggml/1550)

ggerganov committed 2 days ago

974db58c

vulkan: Workaround compiler bug in conv2d coopmat2 path (llama/24924)

jeffbolznv committed 2 days ago

df03e266

CUDA: add cublasSgemmBatched mapping for HIP/MUSA vendor headers (llama/25033)

leonardHONG committed 2 days ago

b564d536

ggml-cpu: fix SVE leftover path in ggml_vec_dot_f32 (llama/24699)

tdakhran committed 2 days ago

5e7ef7e4

sycl : clamp softmax input to avoid underflow (llama/24941)

Jassieluo committed 2 days ago

02be8712

CUDA: batch out_prod broadcast (dps2>1) path with cublasSgemmBatched (llama/24426)

leonardHONG committed 2 days ago

96996579

opencl: flush profiling batch at shutdown for incomplete batches (llama/25016)

shaofeiqi committed 2 days ago

7ce3616a

CUDA: Various fixes to `cpy.cu` (llama/25000)

ORippler committed 2 days ago

9cad7959

ggml : address integer overflows in binary ops CUDA implementation (llama/24706)

fairydreaming committed 2 days ago

82686b51

sycl : support --split-mode tensor (llama/24152)

Spruill-1 committed 2 days ago

ffd5b0b5

sycl : fix the failed UT cases of conv_3d (llama/24900)

arthw committed 2 days ago

c4ba5e2e

opencl: support non-contig rows in norm (llama/24965)

lhez committed 2 days ago

fe37d2e4

hexagon: MUL_MAT and MUL_MAT_ID rework : 32x32 tiled weight repack, kernel-params, cached graphs (llama/24954)

max-krasnyansky committed 2 days ago

e0a0f125

vulkan: allow reducing the graph submission batches to avoid timeouts (llama/24872)

wbruna committed 2 days ago

a201c205

vulkan: fail the build when a shader fails to compile (llama/24450)

liminfei-amd committed 2 days ago

f081d84a

vulkan: Apply bias before softmax in FA, to avoid overflow (llama/24909)

jeffbolznv committed 2 days ago

792da0ec

vulkan: support all backend tests for SQR/SQRT/SIN/COS/CLAMP/LEAKY_RELU/NORM (llama/24582)

jeffbolznv committed 2 days ago

3fc2c57a

vulkan: Support GET_ROWS_BACK (llama/24883)

jeffbolznv committed 2 days ago

1fdbfb19

vulkan: support CONV_3D (llama/24612)

jeffbolznv committed 2 days ago

c25eb26c

vulkan: make mul_mm ALIGNED a spec constant (llama/24689)

jeffbolznv committed 2 days ago

cc0d00f6

vulkan: link ggml-cpu when GGML_VULKAN_CHECK_RESULTS / RUN_TESTS are enabled (llama/24444)

Detensable committed 2 days ago

e3c18388

ggml-webgpu: improve MTP inference by using mat-vec path for small batches (llama/24811)

yomaytk committed 2 days ago

2955e84e

opencl: q8_0 gemv precision improvement (llama/24923)

shawngu-quic committed 2 days ago

0badff27

support bf16 on bin_bcast OP and unary OPs (llama/24838)

arthw committed 2 days ago

1b3f8434

fix(hexagon): use padded stride for ssm-conv weights (llama/24470)

BiReRa committed 2 days ago

556d900c

ggml : optimize AMX (llama/24806)

angt committed 2 days ago

420518d8

ggml-webgpu: add adapter toggles for F16 on Vulkan + NVIDIA

yomaytk committed 2 days ago

75b90fb0

mtmd, arg: fix utf8 handling on windows (llama/24779)

ngxson committed 2 days ago

adbcfa0d

Older