- Notifications
You must be signed in to change notification settings - Fork 32.6k
Pull requests: huggingface/transformers
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
fix: implement Mxfp4Dequantize.reverse_op for save_pretrained support
#44983 opened Mar 25, 2026 by Hyungkeun-Park-Nota Loading…
Trainer: set skip_logits for loss-only eval when liger enabled
#44981 opened Mar 25, 2026 by AkshajKashyap Loading…
4 of 6 tasks
bug-fix: do not assume torch.cuda is available when setting up norm values, even if flash linear attention is available
#44980 opened Mar 24, 2026 by kallewoof Loading…
2 of 6 tasks
fix: handle absent sys.modules entry in modeling_utils
#44978 opened Mar 24, 2026 by cjkindel Loading…
2 of 6 tasks
Refactor core_model_loading to support FSDP shard-on-read loading
#44974 opened Mar 24, 2026 by 3outeille Loading…
Fix max_seqlen type in vision attention for torch.compile + FA2
#44973 opened Mar 24, 2026 by andylizf Loading…
Fix CPU 16 bytes alignment issue using equivalent fallback
#44970 opened Mar 24, 2026 by IlyasMoutawwakil Loading…
6 tasks
Fix FA kernel launch needs correct cuda device ctx in multi-gpu env
#44967 opened Mar 24, 2026 by Qubitium Loading…
2 of 6 tasks
fixed import error with PILImageResampling
#44958 opened Mar 23, 2026 by josh-kean Loading…
3 tasks done
Fix: Add correct return behaviour when output_hidden_states=True for CLIP and SIGLIP vision models
#44952 opened Mar 23, 2026 by Jess-Co-Del Loading…
2 tasks done
feat: Add router_logits override to enable Routing Replay for MoE models
#44951 opened Mar 23, 2026 by hemantmm Loading…
1 task done
Fix: NotebookProgressCallback crash when evaluating with the Trainer
#44949 opened Mar 23, 2026 by Charly21r Loading…
3 tasks done
fix tie_weights skipping logic is not tied to model thread scope
#44940 opened Mar 23, 2026 by Qubitium Loading…
2 of 6 tasks
fix: avoid unconditional model_info call in _patch_mistral_regex
#44923 opened Mar 22, 2026 by prakhar-agarwal • Draft
fix: skip
clean_up_tokenization for BPE tokenizers in PreTrainedTokenizerFast #44915 opened Mar 21, 2026 by maxsloef-goodfire Loading…
3 of 5 tasks
Remove unnecessary expand_as in get_placeholder_mask across VLMs
#44907 opened Mar 21, 2026 by syncdoth Loading…
7 tasks done
fix(models): Fix Perceiver interpolate_pos_encoding interpolating to the source size
#44899 opened Mar 20, 2026 by harshaljanjani Loading…
3 of 5 tasks
Previous Next
ProTip! Type g i on any issue or pull request to go back to the issue listing page.