deepspeedai / deepspeedai/DeepSpeed

Some Problems with ZeRO Optimizations for FP16 Training

Open
#1,061 4 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

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

Description

When using deepspeed, I configured the FP16 parameter, which reduced the memory of my large model training graphics. However, after adding zero's second-stage function, the memory was exceeded. Why? Can't these two functions be used at the same time?

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.

Research direction

No files or tests are identified. Start by reproducing the reported memory increase with FP16 training and ZeRO stage 2, then compare the two configurations and document whether they can be combined and what causes the excess memory; done means a confirmed explanation or a narrowly specified fix.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
distributed-systems, machine-learning, performance
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.