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
- `load_state()` and `save_state()` raise a bare `IndexError` under `automatic_checkpoint_naming`
- Fix VRAM leak from cached int8 weights during CPU/disk offload
- Fix ParallelismConfig.to_json returning None instead of the config dict
- Only disable the compiler on the offload path in AlignDevicesHook
- Keep non-tensor values when concatenating batches
- Keep the batch sampler when preparing on one process
- Saving a configuration mutates its Enum values and empty dictionaries
- Preserve configuration values during serialization
- Use actual GPU ids for balanced_low_0
- Fix disabled ZeRO-3 initialization contexts
- Docs
- Python not yet supported