huggingface/transformers

Pull Requests Commits

Merge branch 'main' into modular_playground

itazap committed 23 days ago

Verified b9c97ce9

fix(moe): Handle dtype mismatch in torch._grouped_mm with autocast (#43839)

Mr-Neutr0n committed 23 days ago

Verified 007bb8c3

persimmon modular

itazap committed 23 days ago

fd839fb4

Train sampler unification (#43138)

jiosephlee committed 23 days ago

Verified 372c27e7

revert place_model_on_device to property (#43895)

SunMarc committed 23 days ago

Verified 8efd8151

Decorate cache updates with no_grad, just in case (#43897)

Rocketknight1 committed 23 days ago

Verified 6c4710bd

Update KERNELS_MIN_VERSION to 0.10.2 to be the same as setup.py (#43753)

cyyever committed 23 days ago

Verified a316807d

Fix typo in quantization_operations in PEFT integrations (#43821)

redpanda1995 committed 23 days ago

Verified 42dc276c

adding BC for custom toks accessing slow tok attrs deprecated in v5 (#43898)

itazap committed 24 days ago

Verified e2372ffa

Minor changes trainer (#43744)

SunMarc committed 24 days ago

Verified 24c743e3

Fix expected DAC outputs due to (old) change in CI settings. (#43896)

ebezzam committed 24 days ago

Verified cc88192b

Remove mask slicing in all eager attentions (#42186)

Cyrilvallez committed 24 days ago

Verified e00bac44

Separate `check_model_inputs` into `capture_outputs` and `merge_with_config_defaults` + ensure correctness (#43862)

Cyrilvallez committed 24 days ago

Verified 42791a34

[fix] DAC model: Apply STE in Dac.from_latents to match the forward pass (#43820)

harshaljanjani committed 24 days ago

Verified 3d24ffaa

itazap committed 24 days ago

84e8cced

Update hub metadata (#43892)

zucchini-nlp committed 24 days ago

Verified 15bada75

[Docs] Add docs for GLM-OCR and fix EomT-DINOv3 (#43710)

NielsRogge committed 24 days ago

Verified 60a5dbaa

[`fix`] Use `last_hidden_state` key from `get_image_features` for llama4 (#43882)

tomaarsen committed 24 days ago

Verified 71d7fc58

Refactor trainer init (#43807)

SunMarc committed 24 days ago

Verified e12aa2d2

fix fbgemm fp8 multi-device load failure. (#43581)

sywangyi committed 24 days ago

Verified 6d86e9a3

fix the error of tests/quantization/fbgemm_fp8/test_fbgemm_fp8.py::Fb… (#43547)

sywangyi committed 24 days ago

Verified b33235e3

[Trainer] Move optimizer cls init to trainer_optimizer.py (#43738)

SunMarc committed 24 days ago

Verified 760eae30

Make it possible to evaluate when using sequence parallel in HF Trainer (#43517)

jp1924 committed 24 days ago

Verified af496d74

rm redundant models

itazap committed 24 days ago

6f240081

Remove unnecessary code or checks for PT 2.4+ (#43787)

cyyever committed 24 days ago

Verified 520fad98

Update TrainingArguments (#43806)

SunMarc committed 24 days ago

Verified 44f92b63

model class match summary

itazap committed 24 days ago

1511e704

Fix old tech stack in doc (#43879)

cyyever committed 24 days ago

Verified 6028dd2b

Fix SwanLab callback to forward resume init args (#43848)

OiPunk committed 24 days ago

Verified 6df88fab

[`Jamba`] Fallback to slow path and warn instead of error out (#43889)

vasqu committed 24 days ago

Verified 9b4de431

Older