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 supported21 Subscribers
Add a CodeTriage badge to deepspeed
Help out
- Issues
- ZeRO-1/2: expose original parameter layout on flat optimizer partitions
- ZeRO-3: partition parameters of modules added after initialization
- [BUG] zero_quantized_nontrainable_weights corrupts frozen weights and crashes ZeRO-3 fetches
- Fix quantized storage of nontrainable weights in ZeRO-3
- Fix comms logger IndexError when the tensor is passed by keyword
- ZeRO-3 NVMe offload: restore the optimizer state on resume
- feat(comm): one-shot P2P custom all-reduce for TP=2 inference
- Keep --venv_script out of the pdsh pkill pattern
- Propagate the user script's exit code from launcher_helper
- fp16 `hysteresis: 0` (or negative) makes DynamicLossScaler never lower the loss scale, so every overflowing step is skipped forever
- Docs
- Python not yet supported