deepspeed
https://github.com/microsoft/deepspeed
Python
DeepSpeed is a deep learning optimization library that makes distributed training easy, efficient, and effective.
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 supported16 Subscribers
Add a CodeTriage badge to deepspeed
Help out
- Issues
- [Bug] Missing input validation for fp16 loss_scale accepts float('inf')
- Fix count_used_parameters_in_backward crash on PyTorch < 2.3 (#7756)
- [BUG] Fix: Fix gradient norm calculation and dynamic shape blocking in PP+ZeRO1 collective communication
- [BUG] DeepSpeed ZeRO-3 deadlock in engine.step() at step 0 under 2-GPU execution (RTX 3090, torch 2.2.1, DS 0.14.2)
- [BUG] Multiple missing zero-guards cause ZeroDivisionError / non-finite values across DeepSpeed (4 locations)
- [BUG] DistributedAttention produces numerically incorrect results vs its own SeqAllToAll-equivalent baseline when sequence length is not divisible by world size
- [BUG] Ulysses crashes in Stage 0
- [REQUEST] Ulysses support for VLMs
- [BUG] Delayed all_gather memory release using ZERO3
- Perform expert parallel inference on deepseek-r1-671b
- Docs
- Python not yet supported