ggerganov/llama.cpp

Pull Requests Commits

cuda: reset device in get_memory function if no backend is active

0cc4m committed 1 hour ago

e21ad065

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

0cc4m committed 3 hours ago

Verified 6e093b80

ci : fix s390x release job (#23898)

ggerganov committed 5 hours ago

Verified 33752857

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

ggerganov committed 6 hours ago

Verified d4204b03

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

rgerganov committed 7 hours ago

Verified 1738129b

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

ngxson committed 14 hours ago

Verified 0821c5fc

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

reeselevine committed 14 hours ago

Verified 151f3a98

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

reeselevine committed 15 hours ago

Verified b22da258

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

ruixiang63 committed 15 hours ago

Verified 689a9a47

app: add llama update self updater (#23865)

ServeurpersoCom committed 15 hours ago

Verified 5a46b46a

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

ValdikSS committed 16 hours ago

Verified 22d66b56

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

tdakhran committed 17 hours ago

Verified 2084434e

graph : ensure DS32 kq_mask_lid is F32 (#23864)

CISC committed 18 hours ago

Verified 764f1e64

server: remove obsolete scripts (#23870)

ngxson committed 18 hours ago

Verified b5f52280

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

ggerganov committed 18 hours ago

Verified dc71236b

download: add option to skip_download (#23059)

ngxson committed 21 hours ago

Verified 06d26dfd

mtmd: Add DeepSeekOCR 2 Support (#20975)

sfallah committed 22 hours 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 1 day ago

Verified cb47092b

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

fairydreaming committed 1 day ago

Verified 1f0aa2a6

llama: use f16 mask for FA to save VRAM (#23764)

am17an committed 1 day ago

Verified 031ddb2e

ggerganov committed 1 day ago

fe12e422

ggml : bump version to 0.13.1 (ggml/1523)

ggerganov committed 1 day ago

ea02bc37

ngram-mod : Add missing include (#23857)

oazizi000 committed 1 day ago

Verified b000431a

llama: add llm_graph_input_mtp (#23643)

am17an committed 1 day ago

Verified eef59a76

app : move licences to llama-app (#23824)

angt committed 1 day ago

Verified 98e480a3

cuda : disables launch_fattn PDL enrollment due to compiler bug (#23825)

aendk committed 1 day ago

Verified 241cbd41

meta : Add missing `buffer` set in allreduce fallback !COMPUTE clear (#23480)

TheBlueMatt committed 1 day ago

Verified 33c718db

hexagon: basic/generic op fusion support and RMS_NORM+MUL fusion (#23835)

max-krasnyansky committed 1 day ago

Verified 19e92c33

mtmd-debug: add color and rainbow mode (#23829)

ngxson committed 1 day ago

Verified 751ebd17

Older