ggml-org/ggml

Pull Requests Commits

Fix CPU backend deadlock in RWKV_WKV6 and GLA with 33+ threads

Copilot committed 99 days ago

aeb817eb

Copilot committed 99 days ago

2f8bceb7

ggml : bump version to 0.9.9 (#1449)

ggerganov committed 100 days ago

Verified a04eea07

sync : whisper.cpp

ggerganov committed 100 days ago

Verified 15547928

sync : llama.cpp

ggerganov committed 102 days ago

404fcb9d

vulkan: add noncontiguous GLU support (llama/21081)

0cc4m committed 102 days ago

54468f4b

hexagon: support for IQ4_NL and MXFP4 (llama/21018)

njsyw1997 committed 102 days ago

35ae589f

rpc : proper handling of data pointers to CPU buffers (llama/21030)

rgerganov committed 102 days ago

960449fd

metal : Fix dimension constraint violation in matmul2d descriptor (llama/21048)

lathrys-at committed 102 days ago

ce3a3c97

hip: use fnuz fp8 for conversion on CDNA3 (llama/21040)

IMbackK committed 102 days ago

95bf9bd5

opencl: allow large buffer for adreno (llama/20997)

lhez committed 102 days ago

64384bc7

fix(ggml): correct RISC-V ISA string canonical ordering for RVV in CMake (llama/20888)

ihb2032 committed 102 days ago

cd7931b4

ggml-cuda: Add NVFP4 dp4a kernel (llama/20644)

michaelw9999 committed 102 days ago

c2c7a83d

CUDA & CPU: support F32 kernel type for `CONV_TRANSPOSE_2D` (llama/17094)

AgainstEntropy committed 102 days ago

83d1f963

mtmd: Add DeepSeekOCR Support (llama/17400)

sfallah committed 102 days ago

4a424d3f

llama: fix llama-model-saver (llama/20503)

JohannesGaessler committed 102 days ago

a0e918ce

sycl : fix wrong variable check by assert (llama/20903)

arthw committed 102 days ago

4b23a6d8

metal : add FLOOR, CEIL, ROUND, TRUNC unary ops (llama/20930)

nuri-yoo committed 102 days ago

19532f0b

metal : add FA instantiations for HSK=512, HSV=512 (llama/20902)

ggerganov committed 102 days ago

80ab0555

hexagon: general DMA and Binary Op fixes for large strides (llama/20918)

max-krasnyansky committed 102 days ago

6ef520c8

opencl: add q6_K gemm and gemv kernels for Adreno (llama/20089)

lhez committed 102 days ago

6b53cddb

rpc : RCE patch (llama/20908)

las7 committed 102 days ago

35b156c9

metal: add CONV_3D (llama/19927)

Ra5hidIslam committed 102 days ago

ee30f1a6

CANN: add RoPE cache preload before ACL graph capture (llama/20747)

noemotiovon committed 102 days ago

e312a9cd

fix(openvino): explicit memset in buffer_context allocation (llama/20857)

thedanhoffman committed 102 days ago

51a1f196

opencl: add flattened Q4_K mv and general Q4_K mm (llama/20773)

shaofeiqi committed 102 days ago

f3239b52

CUDA: fix BF16 FA compilation (llama/20865)

JohannesGaessler committed 102 days ago

fd4e581b

support bf16 and quantized type (llama/20803)

arthw committed 102 days ago

a094f79e

ggml-cuda: native bf16 flash attention for vec kernel (llama/20525)

eous committed 102 days ago

dd1bad2e

Increase number of output elements per-thread block if the K-dimension is small (llama/20635)

gaugarg-nv committed 102 days ago

81419ebe

Older