[ML] Randomize train/test cluster boundary assignment in RDataLoader#22196
Merged
Conversation
Test Results 22 files 22 suites 3d 12h 6m 31s ⏱️ For more details on these failures, see this check. Results for commit 5d6dbf8. ♻️ This comment has been updated with latest results. |
vepadulano
approved these changes
May 15, 2026
Member
vepadulano
left a comment
vepadulano
left a comment
There was a problem hiding this comment.
Nice! See minor considerations from my side.
siliataider
force-pushed
the
rdataloader
branch
2 times, most recently
from
ef1b198 to
87db451
Compare
Member
|
Is a backport to 6.40 needed? |
Contributor
Author
|
/backport to 6.40 |
|
Preparing to backport PR #22196 to branch 6.40 requested by siliataider |
|
This PR has been backported to branch 6.40: #22324 |
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
4 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This Pull request:
Previously
RDataLoaderalways assigned the first fraction of each cluster to training and the last fraction to validation. This meant that across different runs, the train/val split was always identical regardless of the seed.This PR fixes the issue by using the shuffle seed to randomly decide, per cluster, whether training takes the prefix or suffix of that cluster.
This PR fixes #22194