fix: wrap dataset loading in main_process_first to prevent cache race #20
Merged
fix: wrap dataset loading in main_process_first to prevent cache race #20
Conversation
…che races On multi-node runs, all ranks race to build the HF datasets cache simultaneously on Lustre, causing corruption or crashes. Wrapping load_and_mix_datasets() in PartialState().main_process_first() lets rank 0 populate the cache first; all other ranks then read from it.
Neonkraft
changed the title
KonstiNik
reviewed
Apr 30, 2026
Collaborator
KonstiNik
left a comment
KonstiNik
left a comment
There was a problem hiding this comment.
LGTM. Very important addition in my opinion.
I believe it's debatable whether this is a feature or a bug fix.
KonstiNik
approved these changes
Apr 30, 2026
Neonkraft
changed the title
Collaborator
Author
Fair point, this is more of a fix. |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
On multi-node runs, all ranks try to build the HF datasets cache simultaneously on Lustre. Because Lustre lacks the local-filesystem locking semantics that the HF datasets library assumes, concurrent writes corrupt the cache or crash the job. Wrapping
load_and_mix_datasets()inPartialState().main_process_first()ensures rank 0 populates the cache first; all other ranks then read from it.Type of change