ggerganov/llama.cpp

Pull Requests Commits

model : make Gemma 4 shared-KV tail attn_k tensors optional on load (#21739)

MoonRide303 committed 35 days ago

Verified e62fa13c

ggml-webgpu: support non-square subgroup matrix configs for Intel GPUs (#21669)

SharmaRithik committed 35 days ago

Verified bfd1f453

ggml-webgpu: address quantization precision and backend lifecycle managment (#21521)

Constannnnnt committed 35 days ago

Verified e4fed9d0

server : ignore --alias when using --models-preset (#21380)

angt committed 35 days ago

Verified 5dd10253

common : fix when loading a cached HF models with unavailable API (#21670)

angt committed 35 days ago

Verified fb38d6f2

common: mark --split-mode tensor as experimental (#21684)

JohannesGaessler committed 35 days ago

Verified 0893f50f

webui: Static build output improvements (#21667)

allozaur committed 35 days ago

Verified f989a6e3

common : enable reasoning budget sampler for gemma4 (#21697)

berkidem committed 35 days ago

Verified d7ff074c

docs : fix broken link to ggml-openvino in OPENVINO.md (#21709)

ibelem committed 35 days ago

Verified 3f8752b5

vulkan: Support Q1_0 (#21539)

jeffbolznv committed 35 days ago

Verified 7b691253

common : add fluidity to the progress bar (#21671)

angt committed 35 days ago

Verified e095a482

CUDA: fuse muls (#21665)

am17an committed 36 days ago

Verified e34f0421

HIP: add CDNA4 (gfx950) architecture support for MI350X/MI355X (#21570)

andyluo7 committed 36 days ago

Verified d132f22f

ggml: backend-agnostic tensor parallelism (experimental) (#19378)

JohannesGaessler committed 36 days ago

Verified d6f30300

ggml : check return value of CUB calls used in argsort and top-k (they all return cudaError_t) (#21676)

fairydreaming committed 36 days ago

Verified 009a1133

requirements : update transformers to 5.5.1 (#21617)

danbev committed 36 days ago

Verified c8ac02fa

webui: add "Send message on Enter" setting (#21577)

mourix committed 36 days ago

Verified 4ef9301e

common : fix ambiguous grammar rule in gemma4 (#21661)

aldehir committed 36 days ago

Verified ddf03c6d

common : simplify autoparser tagged parser rules (#21216)

aldehir committed 36 days ago

Verified 26229755

model: fix multimodal padding token for gemma3n/gemma4 (#21625)

ngxson committed 36 days ago

Verified 057dba33

mtmd: support dots.ocr (#17575)

ngxson committed 36 days ago

Verified 501aeed1

vocab: add gemma4 tokenizer tests, fix edge case (#21534)

pwilkin committed 36 days ago

Verified 0ec191e1

jinja : support ensure_ascii=true, string repetition and int/float self-filtering (#21623)

kwajiehao committed 36 days ago

Verified 243532e5

metal : add missing mm-id specializations for q1_0 (#21662)

ggerganov committed 36 days ago

Verified 5e9c6354

fix: Model Selector choice sync (#21628)

allozaur committed 36 days ago

Verified 9949ad08

server : fix grammar commandline args (#21543)

AUTOMATIC1111 committed 36 days ago

Verified 3ee9da0e

webui: Add option to pre-encode conversation for faster next turns (#21034)

allozaur committed 36 days ago

Verified 75511a8d

sycl : add flash-attn support for head size 512 (#21654)

qnixsynapse committed 36 days ago

Verified b54cb2e3

ci: drop v5 `all:` composition from labeler.yml (#21627)

Marxist-Leninist committed 36 days ago

Verified 8a65a7a8

vulkan: unify type macros to use Vx instead of _VECx (#21605)

0cc4m committed 36 days ago

Verified 8a132faa

Newer Older