deepspeedai / deepspeedai/DeepSpeed
[REQUEST] can we optimize the logic for checkpoint saving
Open
Nobody has claimed this yet.
enhancement
- Dominant language
- Python
- Stars
- 43.1k
- Forks
- 5k
- Avg merge
- 4d 15h
- Merged PRs (30d)
- 112
Description
Is your feature request related to a problem? Please describe.
checkpoint saving takes lots of time. It would be great if we can optimize it.
Describe the solution you'd like
- I checked the time cost for this function, which takes 5 minutes for large-scale jobs. Can we remove it?
https://github.com/microsoft/DeepSpeed/blob/3678ee177886bd815c6f77e02416abaac629f76a/deepspeed/runtime/engine.py#L2837 - the following code block also takes lots of time. can we remove the barrier()?
from: https://github.com/microsoft/DeepSpeed/blob/3678ee177886bd815c6f77e02416abaac629f76a/deepspeed/runtime/engine.py#L2811 to
https://github.com/microsoft/DeepSpeed/blob/3678ee177886bd815c6f77e02416abaac629f76a/deepspeed/runtime/engine.py#L2820
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start in deepspeed/runtime/engine.py at lines 2811-2820 and 2837, and measure checkpoint-saving time for large-scale jobs before evaluating the proposed removals. Done means the checkpoint path is demonstrably faster while preserving the expected distributed behavior; the issue does not name a test.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- distributed-systems, machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100