All pull requests

Pull requests

• Fix notebook_launcher to propagate mixed_precision to the workers

  #4331·simpleqt opened Sep 24, 2026Contributor·

  ·
  4

• Only disable the compiler on the offload path in AlignDevicesHook

  #4327·jiqing-feng opened Sep 24, 2026Contributor·

  ·
  1

• Fix disabled ZeRO-3 initialization contexts

  #4319·1fanwang opened Sep 23, 2026·

  ·

• Enforce subprocess execution timeouts and clean up interrupted launches

  #4304·Titus-von-Koeller opened Sep 21, 2026Contributor·

  ·
  4

• Fix next-token labels and loss normalization in the accumulation example

  #4303·Titus-von-Koeller opened Sep 21, 2026Contributor·

  ·
  1

• Raise on subprocesses terminated by signals

  #4301·Titus-von-Koeller opened Sep 21, 2026Contributor·

  ·
  1

• Test causal-LM training with DDP, gradient accumulation and mixed precision

  #4300·Titus-von-Koeller opened Sep 21, 2026Contributor·

  ·
  23

• Fix skip_first_batches dropping drop_last, non_blocking, slice_fn, and mesh

  #4296·truecallerabreham opened Sep 20, 2026·

  ·
  1

  3

• Align Accelerator.clip_grad_norm_ with Transformers

  #4266·michaelbenayoun opened Sep 15, 2026Member·

  ·
  3

• Fix compile_regions breaking ModuleDict and ModuleList indexing

  #4260·hjinnkim opened Sep 14, 2026Contributor·

  ·
  2

• Fix shape metadata when copying tensors across devices

  #4245·MrCapricornLiu opened Sep 10, 2026·

  ·

• Fix stateful dataloader checkpoint cache after loading

  #4244·MrCapricornLiu opened Sep 10, 2026·

  ·
  2

• Fix skipped-step detection for AMP-aware optimizers

  #4243·gss10282023 opened Sep 10, 2026·

  ·
  1

  2

• Fix metric padding removal after skip_first_batches

  #4235·Excelius-Wang opened Sep 9, 2026·

  ·

• Add an FSDP2-native Ulysses sequence parallelism backend

  #4209·qgallouedec opened Sep 2, 2026Member·

  ·
  36

• Ask transformers whether a model supports context parallelism

  #4198·qgallouedec opened Aug 31, 2026Member·

  ·
  3

• Allow tensor parallelism next to replicated data parallelism

  #4193·qgallouedec opened Aug 30, 2026Member·

  ·
  1

• Fix stateful dataloader checkpointing across processes

  #4165·jrsmartin opened Aug 14, 2026·

  ·
  6