Skip to content

Pull requests: huggingface/accelerate

Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort

Pull requests list

Stop DataLoaderShard taking a device mesh it never reads
#4252 opened Sep 12, 2026 by vineethsaivs Contributor Loading…
Stop accepting a split_batches argument Accelerator never reads
#4251 opened Sep 12, 2026 by vineethsaivs Contributor Loading…
Do not import torch.distributed in model_has_dtensor on a non-distributed build
#4250 opened Sep 11, 2026 by shoemoney Contributor Loading…
3 of 5 tasks
Restore dataloader state when iteration ends early
#4248 opened Sep 11, 2026 by WenJing95 Loading…
Add callbacks, async checkpoints, and pipeline
#4246 opened Sep 10, 2026 by ak495867 Loading…
5 tasks done
Read ACCELERATE_BYPASS_DEVICE_MAP case-insensitively, like every other boolean flag
#4241 opened Sep 10, 2026 by Sreekant13 Contributor Loading…
2 of 5 tasks
Fix boolean tensor memory accounting
#4236 opened Sep 9, 2026 by Excelius-Wang Loading…
Fix metric padding removal after skip_first_batches
#4235 opened Sep 9, 2026 by Excelius-Wang Loading…
2 tasks done
Resume a mid-epoch checkpoint on the same permutation
#4233 opened Sep 9, 2026 by caiotheodoro Contributor Loading…
4 of 5 tasks
Persist dataloader iteration and shuffle generator in save_state/load_state
#4232 opened Sep 9, 2026 by caiotheodoro Contributor Loading…
4 of 5 tasks
Add an FSDP2-native Ulysses sequence parallelism backend
#4209 opened Sep 2, 2026 by qgallouedec Member Loading…
Fix stateful dataloader checkpointing across processes
#4165 opened Aug 14, 2026 by jrsmartin Loading…
4 of 5 tasks
ProTip! Add no:assignee to see everything that’s not assigned.