ggml-org/ggml

Pull Requests Commits

sync : llama.cpp

ggerganov committed 345 days ago

Verified 8ca55dac

vulkan : add fp16 support for the conv_2d kernel (llama/14872)

Green-Sky committed 345 days ago

Verified 62f0d682

vulkan: skip empty set_rows to avoid invalid API usage (llama/14860)

jeffbolznv committed 345 days ago

Verified fccda47a

Docs: add instructions for adding backends (llama/14889)

am17an committed 345 days ago

Verified 0d380cd7

HIP: Enable Matrix cores for MMQ Kernels, Enable stream-K for CDNA 3 (llama/14624)

deepsek committed 345 days ago

Verified d743740d

CANN: Implement GLU ops (llama/14884)

hipudding committed 345 days ago

Verified ea155558

musa: fix build warnings (unused variable) (llama/14869)

yeahdongcn committed 345 days ago

Verified 81f39973

ggml-cpu : disable GGML_NNPA by default due to instability (llama/14880)

taronaeo committed 345 days ago

Verified a6782ef9

metal: SSM_SCAN performance (llama/14743)

gabe-l-hart committed 345 days ago

Verified afaae7cf

opencl: add fused `rms_norm_mul` (llama/14841)

lhez committed 345 days ago

Verified eab7bdb4

ggml : remove invalid portPos specifiers from dot files (llama/14838)

ORippler committed 345 days ago

Verified 4ad50948

rpc : check for null buffers in get/set/copy tensor endpoints (llama/14868)

struct committed 345 days ago

Verified c7dd462f

sched : fix multiple evaluations of the same graph with pipeline parallelism (llama/14855)

slaren committed 345 days ago

Verified d53fef6a

musa: upgrade musa sdk to rc4.2.0 (llama/14498)

yeahdongcn committed 345 days ago

Verified b8cbf175

contrib : recommend PRs to llama.cpp (#1312)

ggerganov committed 348 days ago

Verified b7bfde9c

cmake : Indent ggml-config.cmake (#1310)

dg0yt committed 349 days ago

Verified 44ea6eda

sync : llama.cpp

ggerganov committed 349 days ago

ac842675

sycl: fixed semantics of block offset calculation (llama/14814)

Alberto Cabrera Pérez committed 349 days ago

3bf67a02

metal : fix fusion across different encoders (llama/14849)

ggerganov committed 349 days ago

4eaa1e3c

sycl: fix undefined variable in work group size check (llama/14843)

djeong20 committed 349 days ago

8b93dd3b

CUDA: fix overflow in FA, tune performance (llama/14840)

JohannesGaessler committed 349 days ago

d589d2d2

CUDA: fix compilation with GGML_CUDA_F16 (llama/14837)

JohannesGaessler committed 349 days ago

e64ed53d

CUDA: fix quantized KV cache + multiple sequences (llama/14822)

JohannesGaessler committed 349 days ago

bf82b878

tests : add non-cont K,V FA tests

ggerganov committed 349 days ago

56c9cd2b

ggml: fix loongarch quantize_row_q8_1 error (llama/14827)

lixing-star committed 349 days ago

4f9b28b6

CANN: weight format to NZ for Ascend310P3 (llama/14407)

tqgy6 committed 349 days ago

010513b8

CUDA: add fused rms norm (llama/14800)

am17an committed 349 days ago

63d8a6ba

vulkan: fix rms_norm_mul to handle broadcasting dim0 (llama/14817)

jeffbolznv committed 349 days ago

bd71636a

cuda : implement bf16 cpy ops and enable bf16 cont (llama/14763)

CISC committed 349 days ago

c32c9dcc

opencl: remove unreachable `return` (llama/14806)

lhez committed 349 days ago

f08563f7

Older