ggerganov/llama.cpp

Pull Requests Commits

opencl: skip the Adreno KQ/KQV image kernels for multi-stream batches (#26189)

wanghqc committed 48 days ago

Verified 8190848b

mtmd : add Nemotron 3 Nano Omni support (parakeet) (#22520)

danbev committed 48 days ago

Verified 7e1e28ca

ui: rendering performance follow-up (#26097)

allozaur committed 48 days ago

Verified 6e2bc65f

docs: Adapt conda-forge package name (#26229)

jjerphan committed 48 days ago

Verified ad77bd31

server: abstract llama_memory calls to common_memory (#26221)

ngxson committed 48 days ago

Verified ee3d1b54

ggml : set output of view src (#25729)

am17an committed 48 days ago

Verified da5b4486

vulkan: add iq4_nl support back to FA (#24585)

jeffbolznv committed 48 days ago

Verified 81616410

ggml-cuda: add chunked SSD matmul for Mamba-2 prefill acceleration (#22675)

BLSharda committed 48 days ago

Verified b62b3509

spec: add DSpark speculative decoding (#25173)

wjinxu committed 48 days ago

Verified 84075273

common/chat: add specialized minimax m3 parser (#26210)

aldehir committed 48 days ago

Verified 6ba5ef24

sycl: fix use-after-return of the SDPA scale in the oneDNN flash-attention path (#25880)

meatposes committed 48 days ago

Verified d6b61ac0

server : add extra trace log for prompt similarity (#26218)

ggerganov committed 48 days ago

Verified 9a3bf2b8

ggml-metal: FWHT kernel for metal backend (#25924)

nicklafleur committed 48 days ago

Verified f95de977

spec: add eagle3-v3 support for gpt-oss model (#25794)

ruixiang63 committed 48 days ago

Verified f8706784

contrib : add guideline about the "merge ready" label (#26178)

ggerganov committed 48 days ago

Verified c6292cfb

Disable -ffast-math on HIP (#25495)

Beinsezii committed 48 days ago

Verified 91f8c9c5

mtmd: support MiMo-V2.5 audio input (RVQ-based model) (#26190)

ngxson committed 48 days ago

Verified 1cbfd198

common : add common_print_available_devices() (#26170)

angt committed 49 days ago

Verified 0e4a0362

model: Add support for Nanbeige4.2 (#25994)

zqlcode committed 49 days ago

Verified b77d6467

fit : count nextn (MTP) blocks in n_gpu_layers so front layers stay on GPU (#26177)

John-194 committed 49 days ago

Verified 0324696b

sycl(build): parallelize ocloc invocations (#25903)

Titaniumtown committed 49 days ago

Verified 8e8681e0

ggml : adjust logic for offloading ops to weight's backend (#25832)

ggerganov committed 49 days ago

Verified dee2a846

tests : remove unnecessary sync in test-save-load-state (#26166)

ggerganov committed 49 days ago

Verified 7ef790f9

common: fix explicit -md precedence over draft sidecar resolution (#26165)

ServeurpersoCom committed 49 days ago

Verified ddfc2288

docs: add exception about weight folding (#26168)

ngxson committed 49 days ago

Verified 419b881c

ggml-cpu: Enable BF16 tiled gemm optimization on PowerPC (#26068)

shalinib-ibm committed 49 days ago

Verified b9102008

args: add `-lm mlock` where it mlocks but doesnt mmap (#26135)

taronaeo committed 49 days ago

Verified ad256ded

server + ui: fix stream routes for model names containing a slash (#26137)

ServeurpersoCom committed 49 days ago

Verified d73c1d6b

ui: Fix symbolic math tool JS sandbox prompt (#26131)

rankaiyx committed 49 days ago

Verified 88b47a75

mtmd: Add Vision Support for Minimax-M3 (#25113)

timkhronos committed 49 days ago

Verified 3d1c3a89

Newer Older