ggml-org/llama.cpp

Pull Requests Commits

ggerganov committed 17 days ago

Verified 2ab78db3

ggerganov committed 18 days ago

Verified d493bc89

cont : helper structs

ggerganov committed 18 days ago

Verified 8e9ed92b

cont : count input tensors before reserving the sched

ggerganov committed 18 days ago

Verified aa75b9f5

llama-context : report graph inputs and input tensors during sched reserve

ggerganov committed 18 days ago

Verified 4b92271f

server : support slot save/restore with media inputs (#26640)

CHIPMUNK-T0T committed 18 days ago

Verified 5d9e5ac3

ui: add read_media tool (#25877)

parabelboi committed 18 days ago

Verified 4dd12758

opencl: default FA c8 cluster width to 16 on X1E (#26433)

wanghqc committed 18 days ago

Verified 89e0aa6f

tests : update speculative params (#26925)

ggerganov committed 18 days ago

Verified a4a4c51f

vulkan: add TQ2_0 (ternary) support (#25850)

michaeltrabalka-tech committed 18 days ago

Verified a7cd2f0e

wavtokenizer-dec : bound posnet/convnext block_count against n_layer_all (#26892)

oakkaya committed 18 days ago

Verified 55f453b9

convert : handle per_layer_config in Gemma4 (transformers 5.15) (#26882)

pluvium27 committed 18 days ago

Verified 6eff5932

opencl: use flat mv q5_k when weight exceeds image1d_buffer_t limit (#26880)

lhez committed 18 days ago

Verified cb27fe9c

chat : fix muse-glimmer detection of tool calls after EOM (#26879)

ruanslv committed 18 days ago

Verified 0b1bad14

ci : add missing release check (#26923)

CISC committed 19 days ago

Verified 7b13a840

CUDA: only disable CUDA graphs when mul_mat_id actually needs a stream sync (#26802)

grafail committed 19 days ago

Verified ebb546b7

cuda : add warp-per-row wkv7 kernel for single-token decode (#26111)

123123213weqw committed 19 days ago

Verified 59886331

spec : update speculative-simple (#26904)

ggerganov committed 19 days ago

Verified f785fc9e

chat : tighten bare function parsing for Qwen models (#26793)

aldehir committed 19 days ago

Verified ba360efe

ci : add windows-rocm to check-release (#26897)

CISC committed 19 days ago

Verified 70dfba5a

imatrix.cpp: Move finite check and only check touched experts (#26861)

bartowski1182 committed 19 days ago

Verified 38406d59

requirements: use stable torch packages on s390x (#26864)

nikwen committed 19 days ago

Verified 2468576f

convert : keep quantization scales for nemotron --mtp export (#26903)

ynankani committed 19 days ago

Verified 5d16e81d

Dflash support for nemotron-3.5 (#26905)

lnigam committed 19 days ago

Verified cc078b45

mtmd: support pocket-tts (#26871)

ngxson committed 19 days ago

Verified 6e62ba53

ui: fix context gauge for single-model usage (#25738)

intel00000 committed 19 days ago

Verified 8d274dd7

ci: hip-quality-check: update vgpr spill ignore list (#26859)

IMbackK committed 19 days ago

Verified 70448594

model-conversion : use save_output_data for causual embeddings [no ci] (#26890)

danbev committed 19 days ago

Verified 1138b851

tests : fix running server tests on windows (#26889)

ggerganov committed 19 days ago

Verified 9afff1b7

llama: add default load-mode auto, which avoids mmap on iGPUs (#26081)

0cc4m committed 19 days ago

Verified 153d324b

Older