[GRPO] Make dataloader deterministic #609
Open
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
meta-cla
bot
added
the
CLA Signed
Contributor
|
LGTM. can you verify that this makes the dataloader deterministic beyond the first epoch too? you can test it by truncating the dataset for example. |
felipemello1
reviewed
Dec 5, 2025
| f"Dataset epoch {self._epoch - 1} completed. Starting epoch {self._epoch}" | ||
| ) | ||
| self._base_dataset.set_epoch(self._epoch) | ||
| self._base_dataset.set_epoch(self._epoch) # for determinism |
Contributor
There was a problem hiding this comment.
Suggested change
| self._base_dataset.set_epoch(self._epoch) # for determinism | |
| self._base_dataset.set_epoch(self._epoch) |
felipemello1
reviewed
Dec 5, 2025
| self._base_dataset = self._base_dataset.map(gsm8k_transform) | ||
| self._base_dataset = self._base_dataset.shuffle() | ||
| self._base_dataset = self._base_dataset.shuffle(seed=self.seed) | ||
| self._base_dataset.set_epoch(self._epoch) # for determinism |
Contributor
There was a problem hiding this comment.
Suggested change
| self._base_dataset.set_epoch(self._epoch) # for determinism | |
| self._base_dataset.set_epoch(self._epoch) |
Contributor
There was a problem hiding this comment.
i suggest we remove the "for determinism". It does a bit more than that: it reshuffles on every epoch, and the epoch number works as a seed
felipemello1
approved these changes
Dec 5, 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Outdated suggestions cannot be applied.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.
This will make the dataloader deliver the data in the same order each time.
Test Plan
Added a breakpoint, and printed out the
promptin pdb:Before,
promptwas different across runs; after this PRpromptis the same across runs.