huggingface/transformers

Pull Requests Commits

Move MTP out of models into MTPCandidateGenerator

ArthurZucker committed 60 days ago

b8c24d1d

Add Multi-Token Prediction (MTP) inference support

ArthurZucker committed 60 days ago

356dec12

fix: compute auxiliary losses when denoising is disabled in D-FINE (#45601)

Abineshabee committed 60 days ago

Verified 5cf79514

Fix configuration reading and error handling for kernels (#45610)

hmellor committed 60 days ago

Verified 91904ac3

qa: bumped mlinter and allow local override (#45585)

tarekziade committed 60 days ago

57f9936a

Processing Utils: continue when content is a string (#45605)

RyanMullins committed 60 days ago

Verified 1e071b25

SonicMoe (#45433)

IlyasMoutawwakil committed 60 days ago

Verified 533c4e1a

fix transformers + torchao nvfp4 serialization (#45573)

vkuzo committed 60 days ago

Verified 03238980

[AMD CI] Fix expectations for Gemma3n (#45602)

Abdennacer-Badaoui committed 60 days ago

Verified 8e64e533

[docs] multi-turn tool calling (#45554)

stevhliu committed 60 days ago

Verified bd69ed2a

Allow for registered experts from kernels hub (#45577)

winglian committed 60 days ago

Verified 706acf5c

[CB] Changes for long generation (#45530)

remi-or committed 60 days ago

Verified 07e38311

Align latest model attention function dispatch (#45598)

Cyrilvallez committed 60 days ago

Verified 7e435bef

Fix `AttributeError` on `s_aux=None` in `flash_attention_forward` (#45589)

jamesbraza committed 61 days ago

Verified a804c74a

Gemma3n and Gemma4 cannot use rotary kernel (#45564)

Cyrilvallez committed 61 days ago

Verified bc4b3304

do not index past decoded chars with special tokens (#45435)

itazap committed 61 days ago

Verified 74a2a4d0

Update dev version (#45583)

vasqu committed 61 days ago

Verified c71125f1

Add expert parallelism (EP) config support for Qwen3 MoE (#45436)

AmineDiro committed 61 days ago

Verified 7cf4241f

Update torchao usage for XPU and CPU (#45560)

jiqing-feng committed 61 days ago

Verified b00b7c08

[docs] per-request sampling params (#45553)

stevhliu committed 61 days ago

Verified decdbdd0

[`Privacy Filter`] Add model (#45580)

vasqu committed 61 days ago

Verified 7187177f

Add ForSequenceClassification heads for the OLMo family (#45551)

earino committed 61 days ago

Verified 864db665

Add IndexCache support for GLM5 DSA (#45424)

louzongzhi committed 61 days ago

Verified bca7eee6

Fix redundant logic in video processing SmolVLM (#45272)

yonigozlan committed 61 days ago

Verified 56175617

Add /v1/completions endpoint (OpenAI legacy completions API) to `transformers serve` (#44558)

rain-1 committed 61 days ago

Verified 0ea540ef

Fix typos (#45574)

vasqu committed 61 days ago

Verified 8a3b7b04

Updated the image cache for Paddle models according to the latest API (#45562)

zhang-prog committed 61 days ago

Verified 71bb903d

[Model] Add SLANet Model Support (#45532)

zhang-prog committed 61 days ago

Verified 3d779723

refactor(Dots1): drop Dots1MoE override to `pass` (inherits from DSV3 MoE) (#45572)

casinca committed 61 days ago

Verified 8ca32875

perf: avoid recomputing rotary_emb for each layer in some Google and ModernBERT models (#45555)

casinca committed 61 days ago

Verified 628c54c2

Older