Commits huggingface/transformers

Compute MoE load-balancing loss per layer to avoid giant one-hot materialization

qgallouedec committed 5 hours ago

201d5624

Use generator with seed for LengthGroupedSampler in Trainer._get_eval_sampler for deterministic eval order with per_device_eval_batch_size > 1 (#48025)

philipshurpik committed 9 hours ago

Verified 94f09cfe

Enable mlinter findings artifact for inline PR reviews (#48117)

ydshieh committed 13 hours ago

Verified decba1d2

Fix ASR pipeline mono conversion for channels-last audio (fixes #47886) (#47888)

Kayvan-Zahiri committed 13 hours ago

Verified bb8f2353

Fix CpmAnt loading: size lm_head to vocab_size (#48012)

jiqing-feng committed 13 hours ago

Verified e872ca25

Delete old mlinter review comments before posting new ones (#48107)

ydshieh committed 14 hours ago

Verified de795d93

Fix Qwen2.5-Omni / Qwen3-Omni-MoE generation with a compilable cache (#47872)

jiqing-feng committed 15 hours ago

Verified 279dbcf4

[Video] Warn and return all frames when num_frames exceeds total_num_frames (#48074)

carlszk committed 15 hours ago

Verified b67f702d

Accept artifact dir as argument in post_mlinter_review.py (#48106)

ydshieh committed 15 hours ago

Verified 604da442

Improve group images by shape (#47964)

guarin committed 15 hours ago

Verified 0e2ce39f

Fix mlinter artifact path (#48088)

ydshieh committed 15 hours ago

Verified f2ef0b94

[Video] Fix convert_to_rgb channel slicing and alpha blending for RGBA videos (#48053)

Ultron09 committed 17 hours ago

Verified e7e8b7fa

[Whisper] Fix decoder position IDs for left-padded batches in longform generation (#48028)

ydshieh committed 18 hours ago

Verified e12c79c5

[PE] Skip test_sdpa_can_dispatch_on_flash for TimmWrapper-backed models (#48064)

ydshieh committed 1 day ago

Verified f9b76f2d

feat: add nvfp4 quantization (#47883)

drbh committed 1 day ago

Verified 80c66744

fix(pipeline): preserve model.generation_config precedence over pipeline defaults (#47752) (#47953)

nithin42 committed 1 day ago

Verified bc777261

[Gemma3] Update integration test expected values for A10G (#48036)

ydshieh committed 1 day ago

Verified ff09c381

Fallback from 'lanczos' to 'bicubic' when on cuda (#48026)

zucchini-nlp committed 1 day ago

Verified e8018693

[serge] Fix 2 integration tests for model `generation` failing with `import_or_config` (other (2)) (#48061)

sergereview[bot] committed 1 day ago

Verified 2dcc93fe

[serge] Fix 2 integration tests regressed by commit b9090ae58cda (PR #47096) (#48060)

sergereview[bot] committed 1 day ago

Verified 02d9442a

[GPT2] Fix encoder_attention_mask being silently discarded in cross-attention (#47946)

DavidJohnQuinlan committed 1 day ago

Verified e0aeab4c

[Fix] Small FA-related test failures in CB (#47341)

remi-or committed 1 day ago

Verified e4075b9e

fix: honor empty processor_kwargs={} in multimodal pipelines (#48044)

hungnnvidia committed 1 day ago

Verified 606e5010

[CircleCI] Enable CI for private forks, no-op for public repo (#48056)

ydshieh committed 1 day ago

Verified bea0343f

:rotating_light: Leftover processors (#47924)

zucchini-nlp committed 1 day ago

Verified 5068901e

Support `BatchFeature` in length-grouped samplers (#48034)

qgallouedec committed 1 day ago

Verified 24fb8989

Fix EOS for candidate generators (#47931)

pjordanandrsn committed 1 day ago

Verified 5b90fc37

[Gemma3n] Update integration test expected values for A10G + torch 2.13 (#48035)

ydshieh committed 1 day ago

Verified f4715397

fix failed test cases for muse_glimmer (#48011)

kaixuanliu committed 1 day ago

Verified 66634317

Cohere compass tests (#47895)

zucchini-nlp committed 1 day ago

Verified 1863da9c