ggerganov/llama.cpp

Pull Requests Commits

llama-server: fix model params not propagated (#21509)

taronaeo committed 1 hour ago

Verified 69c28f15

unicode : add custom Qwen2 regex handler to fix segfault on long input (#21257)

nhs000 committed 2 hours ago

Verified 0d049d6a

llama: remove per-arch tensor name lists (#21531)

JohannesGaessler committed 2 hours ago

Verified a8ec0df4

server : fix restore for checkpoints with pos_min == 0 (#21510)

ggerganov committed 3 hours ago

Verified e8f50826

ggml : deprecate GGML_OP_ADD1 (#21363)

ggerganov committed 3 hours ago

Verified 22fc7913

ggml: Vulkan build, Linux -- output error string for errno on fork failure (#20868) (#20904)

tomoverlund committed 3 hours ago

Verified 2a619f6f

vulkan: add FA dequant for q4_1, q5_0, q5_1, iq4_nl (#21029)

mkoker committed 3 hours ago

Verified edd4d9bc

webui : store reasoning_content so it is sent back in subsequent requests (#21249)

aldehir committed 4 hours ago

Verified 482192f1

ggml-cuda : fix CDNA2 compute capability constant for gfx90a (MI210) (#21519)

aviallon committed 5 hours ago

Verified 71a81f6f

fix: Detect streaming state in reasoning content blocks (#21549)

allozaur committed 5 hours ago

Verified ecce0087

Fix rtl text rendering (#21382)

Kabir08 committed 5 hours ago

Verified d1f82e38

[SYCL] Add Q8_0 reorder optimization (~3x tg speedup on Intel Arc) (#21527)

PMZFX committed 7 hours ago

Verified 0988accf

docs: fix typo in build.md (emdawbwebgpu -> emdawnwebgpu) (#21518)

CastelDazur committed 10 hours ago

Verified 0033f53a

ggml-webgpu: Add the support of `MUL_MAT_ID` (#21147)

yomaytk committed 19 hours ago

Verified d0a6dfeb

ggml: add Q1_0 1-bit quantization support (CPU) (#21273)

khosravipasha committed 20 hours ago

Verified 2e1f0a88

cli: fix stripping of \n in multiline input (#21485)

bipinyadav3175 committed 20 hours ago

Verified 506200cf

[CUDA ] Write an optimized flash_attn_stream_k_fixup kernel (#21159)

gaugarg-nv committed 21 hours ago

Verified 15f786e6

llama-bench: add `-fitc` and `-fitt` to arguments (#21304)

am17an committed 1 day ago

Verified 94ca829b

vocab : add byte token handling to BPE detokenizer for Gemma4 (#21488)

aldehir committed 1 day ago

Verified 4aa962e2

convert : fix block_ff_dim retrieval for lfm2 (#21508)

CISC committed 1 day ago

Verified 941146b3

server : handle unsuccessful sink.write in chunked stream provider (#21478)

lainon1 committed 1 day ago

Verified 482d862b

docs: add hunyuan-ocr gguf, also add test [no ci] (#21490)

ngxson committed 1 day ago

Verified 3979f2bb

convert : set "add bos" == True for Gemma 4 (#21500)

ggerganov committed 1 day ago

Verified 400ac8e1

sycl : handle other FA case (#21377)

arthw committed 1 day ago

Verified f51fd36d

hexagon: slight optimization for argosrt output init (#21463)

YardenTal44 committed 1 day ago

Verified 25eec6f3

llama : correct platform-independent loading of BOOL metadata (#21428)

anchortense committed 1 day ago

Verified 58190cc8

model : add HunyuanOCR support (#21395)

richarddd committed 1 day ago

Verified af76639f

ci : use default RISE RISC-V Runners (#21263)

luhenry committed 1 day ago

Verified 761797ff

server : fix logging of build + system info (#21460)

ddh0 committed 2 days ago

Verified 5d3a4a7d

ci: lower cuda12 floor to 12.8.1 for broader host compatibility (#21438)

M1DNYT3 committed 2 days ago

Verified c08d28d0

Older