Fix random seed on shuffle and interleave_datasets #7823
+56
−0
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
CloseChoice
changed the title
CloseChoice
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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.
closes #7567
Add
shift_rngsmethod toExamplesIterablethat is called directly after sharding. If a generator is available (not the case for all subclasses) we update the seed of the generator by shifting by the worker_id.This is just the fix forshuffle, in the corresponding issueinterleave_datasetsis mentioned as well, which won't be fixed with this approach.EDIT: This is a fix for
shuffleandinterleave_datasets. Adding recursivity toshift_rngssolvedinterleave_datasetsas well. Not sure though if this is completely safe or if we could destroy something with that. I don't think so but could be wrong and appreciate some guidance from the maintainers. I also checked, on a single_worker we are always handing overindex=0so that case preserves the seed the user specified.