Skip to content

Pull requests: huggingface/accelerate

New pull request New
44 Open 2,106 Closed
44 Open 2,106 Closed
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
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Refuse context parallelism for models with sliding-window or chunked attention layers
#4177 opened Aug 20, 2026 by qgallouedec Member Loading…
1
Add FSDP2 activation_checkpointing_offload: offload checkpointed layer inputs to pinned CPU memory
#4175 opened Aug 20, 2026 by qgallouedec Member • 2/2 Loading…
1
Fix circular import between utils.bnb and big_modeling
#4174 opened Aug 20, 2026 by shubhxho Loading…
1
Handle missing accelerate executable in env command
#4168 opened Aug 18, 2026 by tandede Loading…
Forward reduce_batch_size_fn in find_executable_batch_size decorator factory
#4166 opened Aug 16, 2026 by icn5381 Loading…
Fix stateful dataloader checkpointing across processes
#4165 opened Aug 14, 2026 by jrsmartin Loading…
3 of 5 tasks
1
Fix send_to_device crashing on defaultdict batches
#4162 opened Aug 13, 2026 by mayuriphad Loading…
1
2
Fix eval mode for wrapped optimizers
#4161 opened Aug 12, 2026 by uuzzrm Loading…
4 of 8 tasks
1
Forward reduce_batch_size_fn through the find_executable_batch_size decorator form
#4160 opened Aug 11, 2026 by vineethsaivs Loading…
Report MPS as the accelerator in accelerate env
#4158 opened Aug 11, 2026 by xquantize Loading…
2 of 5 tasks
docs: fix documented parameter names that do not match the signature
#4156 opened Aug 10, 2026 by massimiliano1991 Loading…
Preserve defaultdict batches during device placement
#4155 opened Aug 9, 2026 by aswanth-07 Loading…
4 tasks done
1
Add missing type hints in checkpointing.py
#4153 opened Aug 9, 2026 by RudrenduPaul Loading…
Add missing type hints in utils/memory.py
#4152 opened Aug 9, 2026 by RudrenduPaul Loading…
Fix disk offload crash on FP8 tensors
#4151 opened Aug 6, 2026 by shoemoney Loading…
Fix pad_input_tensors to pad the requested dim, not always dim 0
#4150 opened Aug 5, 2026 by shoemoney Loading…
2 of 3 tasks
Fix distributedType -> distributed_type typo in downcast_bf16 guard
#4149 opened Aug 5, 2026 by uttam12331 Loading…
Fix convert_model_to_fp8_ao converting the first and last linear layers
#4147 opened Aug 2, 2026 by vineethsaivs Loading…
3 of 5 tasks
8
[FSDP2] Add additional_modules_to_shard for sharding custom modules
#4145 opened Jul 31, 2026 by yuxinyuan Contributor Loading…
3 of 5 tasks
[docs] Add a concept guide on sharding iterable datasets
#4144 opened Jul 30, 2026 by zokost Loading…
3 of 5 tasks
Add end-to-end distributed test suite (Phase 0: DDP smoke + env parity)
#4143 opened Jul 30, 2026 by Titus-von-Koeller • Draft
1
[FSDP2] Shard eagerly initialized optimizer state
#4142 opened Jul 29, 2026 by sdjasj Loading…
1
[TP] Keep optimizer state aligned after parameter swaps
#4141 opened Jul 29, 2026 by sdjasj Loading…
1
[FSDP] Validate activation checkpointing auto-wrap policy
#4140 opened Jul 29, 2026 by sdjasj Loading…
1
[launch] Report one actionable diagnostic when distributed ranks fail (and make --quiet work for multi-GPU)
#4139 opened Jul 28, 2026 by mosafariuk Loading…
1
ProTip! What’s not been updated in a month: updated:<2026-07-20.