ggerganov/llama.cpp

Pull Requests Commits

metal : remove mul_mm_id amax rescaling

ggerganov committed 1 day ago

e661b09d

mtmd: add state object for mtmd_gen (#30902)

ngxson committed 2 days ago

Verified 0a1ef309

server : forward llama.cpp sampling params on /v1/messages (#30856)

Dgarc359 committed 2 days ago

Verified 261700b8

vendor: update BoringSSL to 0.20261020.0 (#30889)

cabelo committed 2 days ago

Verified f8490953

llama : optimize vocabulary loading (#30679)

mykhailen committed 2 days ago

Verified 824eaecf

llama : place d2t by its set_rows use so it can stay on the GPU (#30912)

am17an committed 2 days ago

Verified eb182d85

metal : fix memory leaks and cleanup logic in GPU capture path (#30890)

nikwen committed 2 days ago

Verified ea9e3446

hexagon: neon-vectorized and threaded repack (tiling) (#30901)

max-krasnyansky committed 2 days ago

Verified 0efe8e4c

opencl: add dp4a GEMM variants to vector load activation (#30753)

lhez committed 2 days ago

Verified b531e4c7

sycl : fix fa crash for Query head dimension mismatch (neq0 != HSK / D=2048/4096) (#30286)

arthw committed 2 days ago

Verified 74b115ca

sycl: collapse duplicate need_check launch bodies in mmq (#30739)

Titaniumtown committed 2 days ago

Verified 96caa800

ggml-cpu : compute f32 gelu_quick and geglu_quick without the fp16 table (#30737)

boshjerns committed 2 days ago

Verified 479a5277

llama : scale moe down projection per expert token (#30861)

ggerganov committed 2 days ago

Verified 8baabd53

Sycl iq3 crash fix (#30604)

anantshri committed 2 days ago

Verified 493253fe

opencl: fix rope program build error with newer Clang (#30893)

bashbaug committed 2 days ago

Verified 339f3242

server : fix usage.completion_tokens undercount with n > 1 (#29486)

unbug committed 2 days ago

Verified 9a3b91c6

server : add POST /tts endpoint for TTS models (#26603)

ngxson committed 2 days ago

Verified 5c0b49a2

server: add realtime transcription API (#30636)

ngxson committed 2 days ago

Verified 1242be66

refactor : move LFM2 shortconv and FFN blocks into a shared LFM base (#30874)

tdakhran committed 2 days ago

Verified 84e66d03

model : support Nemotron-3(.5)-Embed (#30877)

tokinasin committed 2 days ago

Verified 52ffa027

qwen: reshape output to 2d instead of 3d (#30823)

am17an committed 2 days ago

Verified 839a572a

server: Support logprobs with speculative decoding (#27196)

pnb committed 2 days ago

Verified ead9f320

hexagon: fix VTCM size query on images that need layout out-params (#30655)

tdakhran committed 2 days ago

Verified b684b498

mtmd : lfm2 tiles and layout match transformers (#30261)

ykhrustalev committed 2 days ago

Verified 4a45de4f

model : add H2O-Lightning-4B decision model (#30361)

pseudotensor committed 2 days ago

Verified 7e3eeee0

mtmd: fix gemma 4 ffn act (#30882)

ngxson committed 2 days ago

Verified 5ba3b9fd

server : avoid repeated mmproj memory margin addition on sleep/resume (#28233)

TANGBUDU committed 2 days ago

Verified 5a9a34d8

ggml-cuda : support ssm_conv kernel size 7 (#30873)

tdakhran committed 2 days ago

Verified 39da130f

server: Router wasn't propagating the client disconnects fully to the child, the response wait could cause a delayed cancellation (#29707)

Me42019 committed 2 days ago

Verified 4e201388

server: fix missing output_index in Responses API SSE events for compatibility (#30863)

Renzishi committed 2 days ago

Verified 1ba568c5

Older