ggerganov/llama.cpp

Pull Requests Commits

convert: add support for mtp flags

am17an committed 23 days ago

24a84652

convert: fix mtp conversion

am17an committed 23 days ago

bfb5abb7

fix v granularity

am17an committed 23 days ago

96335fc1

llama : fix tensor split for fused qkv with uneven K/V head sizes

am17an committed 24 days ago

c23433c0

CUDA: enable sparse fa for qwen4 (#28770)

am17an committed 27 days ago

Verified 3cf03257

ui: Fix mobile breakpoint + content overflow issues (#29108)

allozaur committed 27 days ago

Verified b23efaa2

fix(mamba) : make time-step projection input contiguous (#28832)

abetlen committed 27 days ago

Verified 42609036

metal: add F16 input to the FWHT (#29094)

bri-prism committed 27 days ago

Verified 9a9f939b

chat : fix gemma4 required tool grammar (#29115)

aldehir committed 27 days ago

Verified f072b103

chat : add dedicated Ling 3.0 (Bailing V3) parser (#28682)

aetherbird committed 27 days ago

Verified 59657a61

hexagon: enable I32 GET_ROWS (#29116)

aparmp-quic committed 27 days ago

Verified e613ef2c

hexagon: add support for GEGLU_QUICK (#29114)

aparmp-quic committed 27 days ago

Verified 851cb34f

hexagon: enable support for TOP_K op (#29113)

aparmp-quic committed 28 days ago

Verified 7d4b92bb

server : improve startup log messages (#29125)

ggerganov committed 28 days ago

Verified 1af554f8

json-schema : accept escaped hyphen in regex patterns (#29127)

ChihebBENCHEIKH1 committed 28 days ago

Verified eb1e1f49

metal : add MoE and SSM_CONV fusion optimizations (#28948)

ggerganov committed 28 days ago

Verified 5b59b83f

metal : fix FA support checks (#29122)

ggerganov committed 28 days ago

Verified 60b06ab9

test-llama-archs : generate dummy test vocab (#29084)

ggerganov committed 28 days ago

Verified efa28e95

metal : support qwen4exp hc ops (#29000)

ggerganov committed 28 days ago

Verified 59fc5a1c

cuda : fix CUB argsort corruption caused by in-place keys (#28389)

TheArchitectit committed 28 days ago

Verified b23701f7

opencl: add support for bin kernel `flash_attn_f32_f16_bin` (#29046)

dsproule committed 28 days ago

Verified 60081bb2

hexagon: add ROLL op support (#29105)

tboinovski1 committed 28 days ago

Verified 2b184703

hexagon: im2col update (#29103)

tboinovski1 committed 28 days ago

Verified 50631b3d

hexagon: HMX flash-attention head_dim padding (support DK=DV=72) (#26539)

tboinovski1 committed 28 days ago

Verified 18a04f09

opencl: add bin kernel `kernel_gemm_noshuffle_q6_k_f32_32b_trans_ila_a8_bin` (#28678)

shaofeiqi committed 28 days ago

Verified ec928150

ggml-cpu: add F16 input to the FWHT (#27779)

bri-prism committed 29 days ago

Verified 4fea119d

ggml : check for allocation failures to prevent crashes (#28149)

sanmai committed 29 days ago

Verified 5b335f41

Model-Saver: Write the SWA pattern, 15 more architectures roundtrip (#29042)

ServeurpersoCom committed 29 days ago

Verified 542348a3

ci: change ubuntu-latest to ubuntu-24.04 (#29079)

taronaeo committed 29 days ago

Verified d663dd3f

ggml-webgpu: fix supports_op condition for GET_ROWS (#28978)

yomaytk committed 29 days ago

Verified 44be98f0

Older