Allow and handle None rewards (miles side)#52
Draft
flukeskywalker wants to merge 1 commit into
Draft
Conversation
flukeskywalker
changed the title
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Sample.remove_sample=Trueas a training-semantic removal, not only a loss-mask change.remove_samplesmetadata and allow zero loss masks only for explicitly removed samples.--rollout-sample-filter-pathhelp/docs and add focused tests for normalization and zero-mask validation.Why
With https://github.com/LLM360/RL360/pull/427, RL360 will mark no-verifier/no-reward Harbor samples with
remove_sample=True. These will havereward=Noneafter https://github.com/LLM360/RL360/pull/415 is merged. Miles needs that marker to mean the sample is excluded from loss as well as training math instead of still influencing group reward baselines or advantage normalization.TODO: make removal of such samples from reward normalization optional, defaulting to False to preserve current behavior.