huggingface/transformers

Pull Requests Commits

Shape cross-attention keys and values by the states they come from

ArthurZucker committed 1 day ago

2109f443

Ignore a stale best checkpoint recorded in a resumed trainer state (#48319)

VaggelisGian committed 1 day ago

Verified 36deb0b5

Keep MXFP4 weights quantized on XPU when use_kernels is set (#47923)

jiqing-feng committed 1 day ago

Verified e15ea9b9

Fix missing FP8 TP layer overrides (#48343)

changwangss committed 1 day ago

Verified d65f37cc

[docs] Fix links and remove TokenizerFast (#47748)

stevhliu committed 1 day ago

Verified 762fae6e

Fix incorrect token classification prefix for ESMC (#48348)

Rocketknight1 committed 1 day ago

Verified 4e5538a3

Create the continuous batching CPU group with local synchronization (#48302)

qgallouedec committed 1 day ago

Verified dabae5fc

Resolve continuous batching config against the text config for composite models (#48299)

qgallouedec committed 1 day ago

Verified 598d8ba8

[Glm 5.3 Flash] GLM 5.3 Flash Support (#48342)

Dovis01 committed 1 day ago

Verified eb4d9e2a

Fix kernel commit and repo paths for ESMFold2 (#48186)

Rocketknight1 committed 1 day ago

Verified ba5cb08d

[`CI`] Unblock fast CI for now (failing tests) (#48344)

vasqu committed 1 day ago

Verified 1e452995

[qwen4_exp] disable torch/onnx export tests due to data-dependent control flow (#48345)

ydshieh committed 1 day ago

Verified 0f0f7e1a

[debug] Trace previous CI run selection in get_previous_daily_ci.py (#48338)

ydshieh committed 1 day ago

Verified 36bc98ef

Normalize HunYuanVL's legacy field aliases via attribute_map (#48261)

hmellor committed 1 day ago

Verified 2c889bf2

Add Qwen4Exp model (#48337)

Cyrilvallez committed 2 days ago

Verified fc5c5bde

Restore BC for the tensor-parallel API (#48300)

ArthurZucker committed 2 days ago

Verified f0e42efb

Fix video-llama modular conversion (#48336)

zucchini-nlp committed 2 days ago

Verified 7d337843

CI: gate the hunyuan-moe slow test (#48330)

tarekziade committed 2 days ago

Verified 949e301b

Add a regression test for force_accelerate_hooks signature preservation (#48260)

wtdcode committed 2 days ago

Verified 5fcf605f

Add an opt-in per-frame pixel cap (cap_pixels_per_frame) to the Qwen3-VL video processor (#48071)

dkrisman committed 2 days ago

Verified 27e7f6c8

Fix `scores` type in stopping criteria docstrings (#47676)

qgallouedec committed 2 days ago

Verified 53e3152d

Docstring check didn't match some file - fix it (#48121)

zucchini-nlp committed 2 days ago

Verified b5416997

Add shared ImageProcessingTester (#47745)

guarin committed 2 days ago

Verified b064ba69

fix bug for blt model parallel bug (#48327)

kaixuanliu committed 2 days ago

Verified 2514a5e9

Cpmant fix use cache (#48013)

jiqing-feng committed 2 days ago

Verified 0f93813a

Fix AutoTokenizer returning TokenizersBackend for DeepSeek-R1-Distill-Qwen models (#48211)

ydshieh committed 2 days ago

Verified 51edaf85

Stop pairing two-word pretokenized documents in the python tokenizer (#48279)

VaggelisGian committed 2 days ago

Verified 6ec0f835

[docs] Fix failing doctests (#47687)

stevhliu committed 2 days ago

Verified 2c1376cf

Fix build_2d_sinusoidal_position_embedding on MPS (#47897)

guarin committed 2 days ago

Verified 562cfd94

Refactor audio buffer creation to use NumPy (#48295)

gsaon committed 2 days ago

Verified efdc97d8

Older