ggml-org/llama.cpp

Pull Requests Commits

readme : add cmd install commands

ggerganov committed 23 days ago

Verified f3e9b9ed

metal : add tensor API flash attention kernel for F16 KV (#29570)

ethanhq committed 23 days ago

Verified 926862e5

llama, server: add /v1/systemone API (models: laya, julia-1, lev, openjev, kev) (#29818)

ngxson committed 23 days ago

Verified a4cb4c61

common : remove fs_open_ifstream() by using u8path() (#29841)

angt committed 23 days ago

Verified 70849ee8

llama : silence unused-result warnings (#29839)

angt committed 23 days ago

Verified 8d81559f

llama : use GGML_ABORT instead of throw (#29840)

angt committed 23 days ago

Verified 6805ae35

opencl: use sigmoid f16 for bf16 (#29787)

lhez committed 23 days ago

Verified a8c9a4e7

SYCL: Q8_0 DMMV ESIMD and MMVQ wide load (#29186)

cwriter committed 23 days ago

Verified 392ded65

vulkan: disable large matmul tile on Samsung GPUs with 32KB shared memory (#28531)

jwsong98 committed 23 days ago

Verified 9e258a6e

sycl: large register file for D=512 FA vec kernels (#29062)

Titaniumtown committed 23 days ago

Verified b9332895

sycl : do not use slow oneDNN reference matmul and fattn (#28985)

lslusarczyk committed 23 days ago

Verified c328acc9

qwen4exp : optimize mask constructions (#29824)

ggerganov committed 23 days ago

Verified 4e2713c1

ggml : add `alloc_buffer_n` to buffer type interface (#23671)

ggerganov committed 23 days ago

Verified 631109b3

ci : fix missing zdnn backend check (#29837)

taronaeo committed 23 days ago

Verified 254b1773

vulkan: add logging to pipeline compile issues (#29794)

0cc4m committed 23 days ago

Verified fb4b2737

pyproject : add linux platform marker to uv torch source (#29177)

Yezat committed 23 days ago

Verified 207bdab9

hexagon: install rebuilt HTP skels (#29828)

kurquhar committed 23 days ago

Verified 5fc4f3c8

qwen4exp: fix tests (#29819)

am17an committed 23 days ago

Verified 159c651f

hexagon: add q2_k and q3_k quant type support (#29717)

jhen0409 committed 23 days ago

Verified a868c3e3

CUDA: fix 2 broken Volta FA cases (#29803)

JohannesGaessler committed 23 days ago

Verified ec7630a6

llama: refer to segment documentation [no ci] (#29074)

JohannesGaessler committed 23 days ago

Verified 78e2964c

common,rpc : fix cache dir creation through symlinks on buggy libstdc++ (#29816)

angt committed 23 days ago

Verified f1cee994

skill: note about model-specific CLI arguments + testings (#29808)

ngxson committed 23 days ago

Verified 68e79bd8

ci: fix Fusion / metal by updating the qwen4exp baseline (#29812)

ServeurpersoCom committed 23 days ago

Verified e358d591

llama : clamp kpool re-pool bound to existing pools (#29805)

ggerganov committed 23 days ago

Verified 81e39ad3

hexagon: shared strided DMA copy for CPY and CONCAT, any-dim CONCAT via DMA (#29685)

njsyw1997 committed 23 days ago

Verified dcd387a4

server: return HTTP 400 for invalid embedding requests (#29060)

SamMalayek committed 23 days ago

Verified d775ebf3

convert : write Gemma embedding scale for DFlash drafts (#29802)

kabu1204 committed 23 days ago

Verified 2b36825c

cuda : route sm70 to the Turing MMVQ nwarps table (#29753)

tkittich committed 23 days ago

Verified 42d95816

metal : release temporary private transfer buffers (#29777)

mvanlamz committed 23 days ago

Verified 13b4d713

Older