ggerganov/llama.cpp

Pull Requests Commits

llama + CUDA: flash_attn_ext_rows (fix unified kv for multi-seq)

am17an committed 11 hours ago

4de4f8fc

server : fix laya abort by limiting n_batch to n_ubatch (#29903)

notforu777 committed 11 hours ago

Verified 1537a0a8

common : add common_is_tty() helper and fix deprecated warnings on Windows (#29860)

angt committed 12 hours ago

Verified edd6e2bb

chat : honor json_schema in Ling 3.0 parser (#29813)

devYRPauli committed 13 hours ago

Verified 9bf55f4a

ci: fix flaky ADD_ADD f16 by using the fused ADD tolerance (#29904)

ServeurpersoCom committed 13 hours ago

Verified a55e952b

graph: gather the recurrent states once so the reserve covers every split (#29856)

ServeurpersoCom committed 14 hours ago

Verified 436f6f89

ggml-openvino: update to 2026.4.1, optimize performance, expand ops, improve device listing. (#29852)

ravi9 committed 17 hours ago

Verified b92761a5

model : Add LFM2.5-Encoder-350M and LFM2.5-Encoder-230M (#29862)

tdakhran committed 19 hours ago

Verified cb7934c5

qwen4exp : halve the indexer score memory (#29825)

ServeurpersoCom committed 21 hours ago

Verified 889edf43

model: add support for clef decision model (text-only) (#29831)

ngxson committed 1 day ago

Verified 99b95488

CUDA: fuse shared experts into MMVQ (#29184)

am17an committed 1 day ago

Verified bed0a856

ci : use t4-medium for cuda jobs (#29842)

CISC committed 1 day ago

Verified 4ebdf2c7

spec : add probabilistic sampling for simple draft and MTP (#27694)

praneshgo committed 1 day ago

Verified 1fb7ef3e

ggml-cuda : fix cpy transposed path corrupting non-contiguous dst (#27663)

devYRPauli committed 1 day ago

Verified 134b2bb7

ggml-quants : avoid invalid rounding in qkx3 scale search (#29817)

devYRPauli committed 1 day ago

Verified 2923cf28

ggml-cpu : fix soft_max_back wrong output when dst aliases src1 (#27096)

devYRPauli committed 1 day ago

Verified dd4c286f

model: support nimble decision model (#29844)

ngxson committed 1 day ago

Verified 46ca246d

readme : add cmd install commands (#29850)

ggerganov committed 1 day ago

Verified d8fbd258

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

ethanhq committed 1 day ago

Verified 926862e5

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

ngxson committed 1 day ago

Verified a4cb4c61

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

angt committed 1 day ago

Verified 70849ee8

llama : silence unused-result warnings (#29839)

angt committed 1 day ago

Verified 8d81559f

llama : use GGML_ABORT instead of throw (#29840)

angt committed 1 day ago

Verified 6805ae35

opencl: use sigmoid f16 for bf16 (#29787)

lhez committed 1 day ago

Verified a8c9a4e7

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

cwriter committed 1 day ago

Verified 392ded65

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

jwsong98 committed 1 day ago

Verified 9e258a6e

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

Titaniumtown committed 1 day ago

Verified b9332895

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

lslusarczyk committed 1 day ago

Verified c328acc9

qwen4exp : optimize mask constructions (#29824)

ggerganov committed 1 day ago

Verified 4e2713c1

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

ggerganov committed 1 day ago

Verified 631109b3

Older