deepspeedai / deepspeedai/DeepSpeed

[Question] why are overlap and contiguous grads meaningless in stage 1 and are ignored

Open
#2,295 6 comments 0 reactions 1 assignee View on GitHub

@tjruwase is already working on this.

Since Dec 12, 2022.

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

Description

https://github.com/microsoft/DeepSpeed/blob/80f94c10c552ec79473775adb8902b210656ed76/deepspeed/runtime/engine.py#L1384

I wonder why we cannot use overlap_comm in zero1 to reduce more latency?
Appreciate any reply.

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.