huggingface/transformers

Pull Requests Commits

Merge branch 'main' into aritra/kernels-updation

MekkCyber committed 1 day ago

Verified 49b394d3

Fix loading of Qwen3 FP8 (#43494)

githubnemo committed 1 day ago

Verified a1f63d55

fix `ShieldGemma2IntegrationTest::test_model` (#43343)

sywangyi committed 1 day ago

Verified 2d5a6d03

Update `SamHQModelIntegrationTest::test_inference_mask_generation_batched_points_batched_images` for `XPU` (#43511)

sywangyi committed 1 day ago

Verified 461cfb14

kernel migration updates for activation

ariG23498 committed 1 day ago

Verified 562d678d

Revert utils files changes from PR #42845 (#43507)

ydshieh committed 1 day ago

Verified dfe30827

Move hardcoded time_step params to config for Bamba, FalconH1, GraniteMoeHybrid (#43461)

raimbekovm committed 1 day ago

Verified fa5dfe06

Prepare inputs for generation is called from `super()` (#43280)

zucchini-nlp committed 1 day ago

Verified 3cbf8471

Enhance repo. consistency bot (#43503)

ydshieh committed 1 day ago

Verified 0189e042

Add `pytest-random-order` for reproducible test randomization (#43483)

tarekziade committed 1 day ago

Verified 9495ae28

Add missing GPURawMetrics.from_dict() method in benchmark_v2 (#43499)

Abdennacer-Badaoui committed 1 day ago

Verified c4b4196e

push dev version 5.0.1.dev0

ArthurZucker committed 1 day ago

6301b503

Fix failing `markuplm` & `perception_lm` integration tests (#43464)

Sai-Suraj-27 committed 1 day ago

Verified 60b9abc8

fix(Phi4Multimodal): Fix incorrect default vision/audio config initialization in Phi4MultimodalConfig (#43480)

charlieJ107 committed 2 days ago

Verified 3abe00a9

handle 1D position_ids for modeling_flash_attention_utils as well (#43403)

kaixuanliu committed 2 days ago

Verified 760946c7

Remove stale TODO comments in UDOP tied weights (#43477)

raimbekovm committed 2 days ago

Verified ab87f244

Fix Mxfp4 dequantize (#43326)

Cyrilvallez committed 2 days ago

Verified d3dfc36b

[`fix`] Sam3Video: Avoid fpn_position_embedding; use pooler_output (#43487)

tomaarsen committed 2 days ago

Verified c8561da8

Revise MIGRATION_GUIDE_V5.md for version 5 updates (#43485)

LysandreJik committed 2 days ago

Verified 64b12e3e

Fix typing hints for different rope parameters per layer type (#43320)

Tcc0403 committed 2 days ago

Verified 59bb7a46

bring back clean_up_tokenization_spaces to tokenizers backend (#43426)

itazap committed 2 days ago

Verified 1e8d1297

Fix lm_head weight tying in Mistral3ForConditionalGeneration (#43429)

aswin00000 committed 2 days ago

Verified 4df39f47

Fix Pixtral with flash attention (#43443)

zucchini-nlp committed 2 days ago

Verified 55d1d093

fix gguf recent conversion issues (#43465)

ArthurZucker committed 3 days ago

Verified a30413b7

43125: Ensure correct checkpoint saving behavior by simplifying Trainer.save_model parallelism logic (#43314)

vaibhav-research committed 3 days ago

Verified 728d64e8

[DeepSpeed] add weight_mapping to _load_state_dict_into_zero3_model (#43303)

kashif committed 3 days ago

Verified f5722a54

fix(models): Add tie_word_embeddings parameter to AyaVisionConfig for proper weight tying (#43455)

harshaljanjani committed 3 days ago

Verified c76c1975

Fix PEFT x MoEs (#43261)

ArthurZucker committed 4 days ago

Verified 3af2eb7a

fix(models): Fix ViVit when dealing with multiple GPUs (#43409)

tarekziade committed 4 days ago

Verified 76ee621a

Fix broken SAM and SAM-HQ due to tie weights and type validation (#43420)

yonigozlan committed 4 days ago

Verified 89b8475d

Older