ggml-org/llama.cpp

Pull Requests Commits

llama-context: add warning if sm tensor is used with one device

am17an committed 54 days ago

921c3e1a

update parameter info and regenerate READMEs

0cc4m committed 54 days ago

a40e4d9b

default mmproj model device to first of global devices value

0cc4m committed 54 days ago

296ab8a9

default draft model devices to global devices value

0cc4m committed 54 days ago

cf1eb320

spec: single device drafter should create meta backend wrapper

am17an committed 56 days ago

477ecf18

sync : ggml (#28379)

ggerganov committed 56 days ago

Verified 64a155d2

model, mtmd: fix gemma4 vision handling (#28335)

ngxson committed 56 days ago

Verified 163a4079

metal : add remaining fa-vec tunings for M3 Max (#28373)

nikwen committed 56 days ago

Verified 8f83678f

ggml : replace compile definitions with version.h.in (#28364)

danbev committed 56 days ago

Verified 86b351fd

Don't use npx inside a package.json script (#28270)

eapache committed 56 days ago

Verified d509cb1e

ggml : don't crash when backend search path can't be read (#28271)

angt committed 56 days ago

Verified 4cbe8b07

ggml : remove GGML_CUDA_PEER_MAX_BATCH_SIZE (#28177)

angt committed 56 days ago

Verified 24f5bf8a

docs : update maintainer PRs link and regenerate AUTHORS (#28365)

ggerganov committed 56 days ago

Verified a529af96

vendor: update BoringSSL to 0.20260903.0 (#28354)

cabelo committed 56 days ago

Verified 38521ec3

ci : disable failing openvino tests (#28347)

ravi9 committed 56 days ago

Verified 0ef4d560

common : make build info output stream configurable (#28322)

angt committed 56 days ago

Verified c390d0ab

ggml-cpu(s390x) : fix q5_1 uninitialized v_acc (#28332)

taronaeo committed 56 days ago

Verified 832fd6f1

src : add n_expert_used_max function (#28323)

danbev committed 56 days ago

Verified 9a4843cf

sycl: fuse rms_norm+mul+add and add+add residual chains (#27610)

newjordan committed 56 days ago

Verified 6703d789

SYCL: Refactor GGML_SYCL_ENABLE_MKL_FA to global var (#26863)

johnkarlhill committed 57 days ago

Verified f9f09f02

llama: fix whole source code rebuilt on each new commit (#28278)

ngxson committed 57 days ago

Verified d230ddd7

common/json-schema : fix GBNF grammar generation for empty object schemas (#28279)

SergeySklyarov committed 57 days ago

Verified c5a5535e

opencl: quant lm_head / decode GEMV and medium-batch GEMM optimizations (speculative decoding/MTP) (#26477)

wanghqc committed 57 days ago

Verified 95ef7fc1

tune MMVQ to MMQ crossover for SM87 (#28285)

kbenkhaled committed 57 days ago

Verified 8c1a2516

snapdragon: ci updates to use new run script (#28293)

max-krasnyansky committed 57 days ago

Verified d30500b8

ops: add Hexagon to ops.md and update main README.md (#28263)

tboinovski1 committed 57 days ago

Verified e107984b

server : use pytest-xdist for server tests (#28298)

danbev committed 57 days ago

Verified 42f0225f

mtmd: propagate const to preproc class (#28310)

ngxson committed 57 days ago

Verified de8656bd

metal : add sparse FA (#28098)

ggerganov committed 57 days ago

Verified 7bb0fc18

metal : fix glu dispatch with ne00 = 1 (#28306)

ggerganov committed 57 days ago

Verified 0df017d6

Older