feat: online reward normalization (Welford’s algorithm)#427
Open
RUFFY-369 wants to merge 4 commits intoNousResearch:mainfrom
Open
feat: online reward normalization (Welford’s algorithm)#427RUFFY-369 wants to merge 4 commits intoNousResearch:mainfrom
RUFFY-369 wants to merge 4 commits intoNousResearch:mainfrom
Conversation
…lity Add RewardNormalizer to atroposlib/envs/ with: - Welford's online algorithm for running mean/variance (no data storage) - Z-score and min-max normalization modes - Configurable reward clipping and warmup period - Checkpoint save/load support - Opt-in integration in BaseEnv via 3 new config fields - WandB metrics for normalization statistics 21/21 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Added an online reward normalizer to
BaseEnvto keep training stable as rewards shift. I used Welford’s Online Algorithm for the running Z-score calculation to keep it O(1) in memory and avoid storing large reward histories.I included a configurable
warmup_stepsphase so the distribution doesn't start shifting until the mean/std estimates have statistically stabilized. This should fix the gradient explosion issues often seen in early RL training stages.Related Issues
Type of Change
✅ Developer & Reviewer Checklist