Pull Requests huggingface/transformers

Fix resized LM head weights being overwritten by post_init

#45079 opened 2026-03-28 00:06 by javierdejesusda

throw error when conversion required

#45078 opened 2026-03-27 23:06 by itazap

fix: pin 50 unpinned actions to commit SHA, extract 1 secret to env var

#45077 opened 2026-03-27 22:20 by dagecko

Osman-Level Innovations: Hardware-Aware Advisor & Selective Weight Surgery CLI

#45076 opened 2026-03-27 20:30 by osman-akkawi

[WIP] Add Deepseek-OCR2 model

#45075 opened 2026-03-27 20:14 by thisisiron

fix(models): Fix dtype mismatch in SwitchTransformers and TimmWrapperModel

#45074 opened 2026-03-27 20:02 by harshaljanjani

Refactor OwlViT to modular Transformers

#45073 opened 2026-03-27 20:00 by Aravind-11

Fix TypeError in rope validation when ignore_keys is a list

#45069 opened 2026-03-27 19:21 by Fr0do

feat: trainer resume_from_checkpoint support hub downloads (#43375)

#45067 opened 2026-03-27 18:26 by aarushisingh04

refactor: shard checkers

#45064 opened 2026-03-27 16:26 by tarekziade

CB improvements for serving

#45063 opened 2026-03-27 16:07 by SunMarc

Fix PIL backend fallback when torchvision is unavailable

#45060 opened 2026-03-27 13:43 by ErenAta16

[serving] Fix continuous batching JSON response serialization

#45057 opened 2026-03-27 13:02 by NathanHB

[`auto_docstring`] needs to be only run on __doc__

#45056 opened 2026-03-27 11:36 by ArthurZucker

Save model config in Trainer checkpoints for non-PreTrainedModel models

#45055 opened 2026-03-27 11:31 by vasanthrpjan1-boop

Fix failing `XCLIPModelIntegrationTest`

#45053 opened 2026-03-27 11:11 by Sai-Suraj-27

Fix NaN weights on non-rank-0 FSDP processes

#45050 opened 2026-03-27 09:19 by albertvillanova

investigate modular conversion speedups

#45046 opened 2026-03-27 08:31 by tarekziade

[Bugfix] Remove incorrect torchvision requirement from PIL backend image processors for patch

#45045 opened 2026-03-27 08:19 by Lidang-Jiang

fix bug for janus model image generation

#45044 opened 2026-03-27 07:50 by kaixuanliu

[3dconv][qwenvl] convert patch embed forward to linear

#45041 opened 2026-03-27 03:50 by JJJYmmm

Llama3 video fix

#45040 opened 2026-03-27 02:56 by sywangyi

add missing colon in custom_attention function signature in attention…

#45037 opened 2026-03-26 23:48 by asuryateja

Pass packed boundary metadata to Qwen3.5 linear-attention fast kernels from data collator

#45034 opened 2026-03-26 20:52 by sdharani91

TP refactor for FSDP + TP integration

#45028 opened 2026-03-26 17:50 by 3outeille

[CB] Add per-request logits processors

#45026 opened 2026-03-26 17:00 by remi-or

Dynamic auto mapping (PoC)

#45018 opened 2026-03-26 11:18 by zucchini-nlp

[WIP][Fix] GLM 5 set `apply_rotary_pos_emb` to `is_neox_style=False` && remove `F.relu()`

#45017 opened 2026-03-26 09:21 by JaredforReal

feature: added import complexity checker

#45013 opened 2026-03-26 07:47 by tarekziade

refactor: added cache in check_repo

#45012 opened 2026-03-26 07:10 by tarekziade