feat: numerical verification for RL distribution health#429
Open
RUFFY-369 wants to merge 4 commits intoNousResearch:mainfrom
Open
feat: numerical verification for RL distribution health#429RUFFY-369 wants to merge 4 commits intoNousResearch:mainfrom
RUFFY-369 wants to merge 4 commits intoNousResearch:mainfrom
Conversation
Add numerical_verification.py to atroposlib/utils/ with: - verify_reward_determinism: N-run reproducibility check - verify_advantage_stability: NaN/Inf/magnitude detection - compare_fp_precision: FP32/FP16/BF16 divergence analysis - verify_score_distribution: collapse/explosion/bias detection - Structured report dataclasses with diagnostic summaries 25/25 tests passing.
for more information, see https://pre-commit.ci
This was referenced Mar 30, 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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
PR Type
📝 General Information
Description
This is a hygiene/safety PR adding
NumericalVerificationutilities. It’s designed to proactively catch Dead Rewards (collapse) or Exploding Advantages (NaNs) in the training loop before they waste valuable GPU hours.It generates a
DistributionReportthat logs and warns if the rewards look biased or collapsed. I’ve integrated this into thewandb_logloop so you can see distribution health directly on your dashboard in real-time.Related Issues
Type of Change
✅ Developer & Reviewer Checklist