ggerganov/llama.cpp

Pull Requests Commits

cont : add kernels/mul_mv_mma.metal

ggerganov committed 33 days ago

9cf70aca

cont : drop Q4_0 ne11=2 special-case

ggerganov committed 33 days ago

17acbc52

cont : clean-up operand/residual logic

ggerganov committed 33 days ago

3af8f157

cont : drop has_tensor gate

ggerganov committed 33 days ago

8c084b0b

cont : clean-up

ggerganov committed 33 days ago

5a993475

Merge remote-tracking branch 'origin/master' into metal-few-row-mma

pratiknarola-t committed 34 days ago

Verified ac369808

metal : remove the CPY_BATCH fusion and the memory range changes

pratiknarola-t committed 34 days ago

Verified e0356f6e

mtmd : fix deprecated strdup warning on Windows (#29863)

angt committed 34 days ago

Verified 836d5717

vendor : update cpp-httplib to 0.59.0 (#29886)

cabelo committed 34 days ago

Verified eec18f5d

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

notforu777 committed 34 days ago

Verified 1537a0a8

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

angt committed 34 days ago

Verified edd6e2bb

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

devYRPauli committed 34 days ago

Verified 9bf55f4a

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

ServeurpersoCom committed 34 days ago

Verified a55e952b

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

ServeurpersoCom committed 34 days ago

Verified 436f6f89

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

ravi9 committed 34 days ago

Verified b92761a5

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

tdakhran committed 34 days ago

Verified cb7934c5

qwen4exp : halve the indexer score memory (#29825)

ServeurpersoCom committed 35 days ago

Verified 889edf43

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

ngxson committed 35 days ago

Verified 99b95488

CUDA: fuse shared experts into MMVQ (#29184)

am17an committed 35 days ago

Verified bed0a856

metal : few-row MMA mat-mul and batched copies for speculative decoding

pratiknarola-t committed 35 days ago

Verified cc470b64

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

CISC committed 35 days ago

Verified 4ebdf2c7

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

praneshgo committed 35 days ago

Verified 1fb7ef3e

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

devYRPauli committed 35 days ago

Verified 134b2bb7

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

devYRPauli committed 35 days ago

Verified 2923cf28

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

devYRPauli committed 35 days ago

Verified dd4c286f

model: support nimble decision model (#29844)

ngxson committed 35 days ago

Verified 46ca246d

readme : add cmd install commands (#29850)

ggerganov committed 35 days ago

Verified d8fbd258

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

ethanhq committed 35 days ago

Verified 926862e5

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

ngxson committed 35 days ago

Verified a4cb4c61

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

angt committed 35 days ago

Verified 70849ee8

Older