ggml-org/llama.cpp

Pull Requests Commits

vulkan: fix check results async upload issue

0cc4m committed 1 hour ago

9cc707ad

llama: only use one iGPU device by default (#23897)

0cc4m committed 2 hours ago

Verified 22cadc19

webui: add custom CSS injection via config (#23904)

ServeurpersoCom committed 11 hours ago

Verified d749821d

Support `-fa auto` in llama-bench (#23714)

gaugarg-nv committed 12 hours ago

Verified aa46bda8

opencl: support bf16 by converting to f16 (#23839)

lhez committed 15 hours ago

Verified d6588daa

ui: exclude generated build dirs from prettier and eslint so lint errors stop being masked (#23910)

ServeurpersoCom committed 18 hours ago

Verified d38d50e7

TP: fix granularity for Qwen 3.5/3.6 + 3 GPUs (#23843)

JohannesGaessler committed 19 hours ago

Verified 8b0e0db6

metal : restore im2col implementation for large kernels (#23901)

ggerganov committed 20 hours ago

Verified 2d9b7c8e

test: (test-llama-archs) log the config name first (#23885)

ngxson committed 22 hours ago

Verified e674b127

ci : update ios-xcode release job to macos-26 (#23906)

ggerganov committed 22 hours ago

Verified 4c4e91b7

ggml : add some lsx support (#23798)

MQ-mengqing committed 1 day ago

Verified d48a56ef

vulkan: add Flash Attention support for BFloat16 KV cache (#23420)

0cc4m committed 1 day ago

Verified 6e093b80

ci : fix s390x release job (#23898)

ggerganov committed 1 day ago

Verified 33752857

ci : clear cache instead of "no timestamp" keys + fix macos (#23895)

ggerganov committed 1 day ago

Verified d4204b03

llama : do not skip iGPU when only RPC devices are present (#23868)

rgerganov committed 1 day ago

Verified 1738129b

server: in SSE mode, send HTTP headers when slot starts (#23884)

ngxson committed 1 day ago

Verified 0821c5fc

ggml-webgpu: Check earlier for WebGPU required features (#23879)

reeselevine committed 1 day ago

Verified 151f3a98

ggml-webgpu: add q4_0/q8_0 SET_ROWS (#23760)

reeselevine committed 1 day ago

Verified b22da258

server-bench : add speed-bench for speculative decoding benchmarking (#23869)

ruixiang63 committed 1 day ago

Verified 689a9a47

app: add llama update self updater (#23865)

ServeurpersoCom committed 1 day ago

Verified 5a46b46a

ui: handle audio/vnd.wave as audio WAV file (#23754)

ValdikSS committed 1 day ago

Verified 22d66b56

vocab : support tokenizer for LFM2.5-8B-A1B (#23826)

tdakhran committed 1 day ago

Verified 2084434e

graph : ensure DS32 kq_mask_lid is F32 (#23864)

CISC committed 1 day ago

Verified 764f1e64

server: remove obsolete scripts (#23870)

ngxson committed 1 day ago

Verified b5f52280

ci : update macos release to use macos-26 runner (#23878)

ggerganov committed 1 day ago

Verified dc71236b

download: add option to skip_download (#23059)

ngxson committed 1 day ago

Verified 06d26dfd

mtmd: Add DeepSeekOCR 2 Support (#20975)

sfallah committed 1 day ago

Verified da3f990a

CUDA: Check PTX version on host side to guard PDL dispatch (#23530)

ORippler committed 1 day ago

Verified 6ed481ee

server: bump timeout to 3600s (#23842)

ngxson committed 2 days ago

Verified cb47092b

model : support for DeepseekV32ForCausalLM with generic DeepSeek Sparse Attention (DSA) implementation (#23346)

fairydreaming committed 2 days ago

Verified 1f0aa2a6

Older