deepspeedai / deepspeedai/DeepSpeed

Can you support automatic mixed precision (amp) in zero optimizer?

Open
#121 2 comments 12 reactions 1 assignee View on GitHub

@xiexbing is already working on this.

Since Sep 8, 2023.

enhancement
Dominant language
Python
Stars
43.1k
Forks
5k
Avg merge
4d 15h
Merged PRs (30d)
112

Description

Currently, the fp16 optimizer and zero optimizer are implementing 'manual' mixed precision algorithm. There are more and more people that are using amp. Nvidia apex is also warning users of its fp16 optimizer (https://github.com/NVIDIA/apex/blob/master/apex/fp16_utils/fp16_optimizer.py#L20).

Can you add an amp-version of mixed precision implementation in zero optimizer? Thanks.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.