ggerganov/llama.cpp

Pull Requests Commits

server : test alternative LRU logic

ggerganov committed 169 days ago

Verified b98f80a6

HIP: add GGML_HIP_MMQ_MFMA option to allow disableing the MFMA path. (#14930)

IMbackK committed 169 days ago

Verified b77d1117

HIP: Ignore unsupported unroll transformation in fattn-vec (#14931)

IMbackK committed 169 days ago

Verified c7aa1364

common : avoid logging partial messages (which can contain broken UTF-8 sequences) (#14937)

kallewoof committed 169 days ago

Verified 1a67fcc3

CANN: Add ggml_set_rows (#14943)

hipudding committed 169 days ago

Verified 204f2cf1

cuda : add softcap fusion (#14907)

CISC committed 169 days ago

Verified 138b288b

server-bench: make seed choice configurable (#14929)

JohannesGaessler committed 169 days ago

Verified bbd0f917

CUDA: add roll (#14919)

am17an committed 169 days ago

Verified 0a5036be

opencl : add ops docs (#14910)

lhez committed 170 days ago

Verified 8ad7b3e6

test-backend-ops : extend test case filtering (#14865)

tlemo committed 170 days ago

Verified bda62193

llama-bench : use local GPUs along with RPC servers (#14917)

rgerganov committed 170 days ago

Verified c556418b

ggml-cpu : deduplicate scalar implementations (#14897)

xctan committed 170 days ago

Verified db16e283

SYCL: Add set_rows support for quantized types (#14883)

qnixsynapse committed 170 days ago

Verified cd1fce6d

mtmd : add support for Voxtral (#14862)

ngxson committed 170 days ago

Verified 00fa15fe

CUDA: fix pointer incrementation in FA (#14916)

JohannesGaessler committed 170 days ago

Verified 946b1f68

model : add support for SmallThinker series (#14898)

wdl339 committed 170 days ago

Verified 6c6e397a

sycl: refactor quantization to q8_1 (#14815)

Alberto Cabrera Pérez committed 170 days ago

Verified afc0e896

ops : update BLAS (#14914)

ggerganov committed 170 days ago

Verified a5771c9e

ops : update Metal (#14912)

ggerganov committed 170 days ago

Verified c35f9eaf

ggerganov committed 170 days ago

1f45f289

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

dg0yt committed 170 days ago

613c5095

quantize : update README.md (#14905)

EAddario committed 171 days ago

Verified 7f975995

vulkan: add ops docs (#14900)

0cc4m committed 171 days ago

Verified bf78f543

SYCL: add ops doc (#14901)

qnixsynapse committed 171 days ago

Verified bbfc8492

llama : clarify comment about pp and tg graphs [no ci] (#14895)

danbev committed 171 days ago

Verified ca0ef2dd

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

Green-Sky committed 171 days ago

Verified 89d10295

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

jeffbolznv committed 171 days ago

Verified f1a4e72d

model : make rope_yarn_log_mul optional for deepseek2 (#14896)

gabriellarson committed 171 days ago

Verified 4762ad73

llama : fix kq_scale for the attention layers of PLaMo2 (#14892)

mitmul committed 171 days ago

Verified 1dc9614e

Docs: add instructions for adding backends (#14889)

am17an committed 172 days ago

Verified 446595b9

Older