accelerate
https://github.com/huggingface/accelerate
Python
🚀 A simple way to train and use PyTorch models with multi-GPU, TPU, mixed-precision
Triage Issues!
When you volunteer to triage issues, you'll receive an email each day with a link to an open issue that needs help in this project. You'll also receive instructions on how to triage issues.
Triage Docs!
Receive a documented method or class from your favorite GitHub repos in your inbox every day. If you're really pro, receive undocumented methods or classes and supercharge your commit history.
Python not yet supported1 Subscribers
Add a CodeTriage badge to accelerate
Help out
- Issues
- Fix tied parameter cache corrupting registered buffer views
- FSDP2: shard a checkpointed layer once, at its checkpoint wrapper
- Fix environment rollback when patch_environment setup fails
- Failed dataloader iteration leaves stale gradient state
- Restore dataloader state after iteration errors
- save_state deletes retained checkpoints before a replacement save succeeds
- Preserve retained checkpoints until save_state succeeds
- fix(modeling): respect dtype when device_map is None
- Fix parameterless modules and repeated forwards in CPU offload hooks
- Preserve RNG state during FSDP2 activation checkpointing
- Docs
- Python not yet supported