Extend invariant suite with gradient-checkpointing equivalence#5689
Merged
Conversation
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
kashif
approved these changes
May 4, 2026
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds
gradient_checkpointing=Falseas a third member of the existingsftanddpoequivalence classes intests/invariant/. GC trades compute for memory and should produce trajectories identical to the canonical (gc=False) within fp32 noise; divergence here would be a real bug in the recompute path.Note
Low Risk
Low risk: changes are limited to the invariant test suite, adding extra configurations to validate trajectory equivalence when
gradient_checkpointingis disabled, with no production code impact.Overview
Extends the invariant equivalence tests for both
sftanddpoby adding a third member configuration that runs withgradient_checkpointing=Falseand asserts its loss/grad-norm trajectory matches the class reference within existing tolerances.Reviewed by Cursor Bugbot for commit faf55fb. Bugbot is set up for automated code reviews on this repo. Configure here.