Change Mlflow monitor process from fork to spawn to reduce memory usage #3830
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
added 4 commits
April 22, 2025 20:07
There was a problem hiding this comment.
Pull Request Overview
This PR resolves a memory hang issue in the MLflow monitor process by switching from fork to spawn, ensuring that the full model memory is released at the correct time.
- Updated process creation to use spawn context via spawn_context.Process.
- Replaced multiprocessing.Event with threading.Event in the run method.
- Revised signal handling by introducing SIGUSR1 and SIGUSR2 for normal and crash exits, respectively.
dakinggg
changed the title
irenedea
approved these changes
Apr 23, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do?
The spawned mlflow monitor process was hanging on to memory for the whole model (initialized in the main process, and not freed until FSDP wrapping, which occurs after the monitor process is started). This resolves that by using spawn instead of fork for the monitor process.
Before and after memory usage:
Large model run that was hanging previously now works:
405b-mlf-after-1-Y26NwDManual test of error run:
70b-mlf-after-2-8OZEl7Statuses all look correct:
