ggerganov/llama.cpp

Pull Requests Commits

metal : per-device tuned (Q, NE) for flash-attn vec (#25750)

forforever73 committed 16 days ago

Verified 96ece714

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

ggerganov committed 17 days ago

Verified 7185fef9

metal: fuse snake activation (mul, sin, sqr, mul, add) (#25459)

ggerganov committed 17 days ago

Verified a9a5d1da

metal : add Q2_0 support (#25419)

ggerganov committed 17 days ago

Verified 70f70ed5

metal : add CONV_2D_DW (depthwise convolution) support (#21565)

ggerganov committed 17 days ago

Verified ccb84453

metal : add set_rows with src0 f16 (#25434)

ggerganov committed 17 days ago

Verified 6d755645

metal: add col2im_1d op (f32/f16/bf16) (#25176)

ggerganov committed 17 days ago

Verified 28944067

metal : per-op source split + parallel compile (#24021)

forforever73 committed 17 days ago

Verified 8ed30a71

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

danbev committed 17 days ago

Verified 7e1e28ca

ui: rendering performance follow-up (#26097)

allozaur committed 17 days ago

Verified 6e2bc65f

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

jjerphan committed 17 days ago

Verified ad77bd31

server: abstract llama_memory calls to common_memory (#26221)

ngxson committed 17 days ago

Verified ee3d1b54

ggml : set output of view src (#25729)

am17an committed 17 days ago

Verified da5b4486

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

jeffbolznv committed 17 days ago

Verified 81616410

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

BLSharda committed 17 days ago

Verified b62b3509

spec: add DSpark speculative decoding (#25173)

wjinxu committed 17 days ago

Verified 84075273

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

aldehir committed 17 days ago

Verified 6ba5ef24

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

meatposes committed 17 days ago

Verified d6b61ac0

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

ggerganov committed 17 days ago

Verified 9a3bf2b8

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

nicklafleur committed 17 days ago

Verified f95de977

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

ruixiang63 committed 17 days ago

Verified f8706784

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

ggerganov committed 17 days ago

Verified c6292cfb

Disable -ffast-math on HIP (#25495)

Beinsezii committed 17 days ago

Verified 91f8c9c5

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

ngxson committed 17 days ago

Verified 1cbfd198

common : add common_print_available_devices() (#26170)

angt committed 18 days ago

Verified 0e4a0362

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

zqlcode committed 18 days ago

Verified b77d6467

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

John-194 committed 18 days ago

Verified 0324696b

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

Titaniumtown committed 18 days ago

Verified 8e8681e0

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

ggerganov committed 18 days ago

Verified dee2a846

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

ggerganov committed 18 days ago

Verified 7ef790f9

Older