manager: Add option to skip initial sync #159
Merged
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
Summary: We currently always heal on step 0 to avoid synchronization issues. We want an option to support skipping this sync for users who set the PyTorch seed so all ranks are initialized with the same values. This diff added a init_sync boolean flag that can be passed from the manager client in python to the manager service in rust. If the manager service skips the sync depending on whether the init_sync is true. Test Plan: Reviewers: Subscribers: Tasks: Tags:
Summary: Test Plan: Reviewers: Subscribers: Tasks: Tags:
Summary: Test Plan: Reviewers: Subscribers: Tasks: Tags:
Summary: Test Plan: Reviewers: Subscribers: Tasks: Tags:
Summary: Test Plan: Reviewers: Subscribers: Tasks: Tags:
facebook-github-bot
added
the
CLA Signed
fduwjj
approved these changes
Apr 8, 2025
WarrenZhu050413
pushed a commit
to guodong/torchft
that referenced
this pull request
Apr 14, 2025
* add option to skip initial sync in Manager Summary: We currently always heal on step 0 to avoid synchronization issues. We want an option to support skipping this sync for users who set the PyTorch seed so all ranks are initialized with the same values. This diff added a init_sync boolean flag that can be passed from the manager client in python to the manager service in rust. If the manager service skips the sync depending on whether the init_sync is true. Test Plan: Reviewers: Subscribers: Tasks: Tags: * Fixed some rust files and tests Summary: Test Plan: Reviewers: Subscribers: Tasks: Tags: * Fix init_sync logic Summary: Test Plan: Reviewers: Subscribers: Tasks: Tags: * Add tests for manager.rs Summary: Test Plan: Reviewers: Subscribers: Tasks: Tags: * Added skip init_sync tests to python client Summary: Test Plan: Reviewers: Subscribers: Tasks: Tags: * manager: final changes for init_sync --------- Co-authored-by: Dave Lei <[email protected]>
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.
This is a version of #127 with a few fixes to support init_sync.
Fixes #117
Test plan: