[train] Add DRO (Direct Reward Optimization) policy loss#1259
Draft
tyler-griggs wants to merge 2 commits intomainfrom
Draft
[train] Add DRO (Direct Reward Optimization) policy loss#1259tyler-griggs wants to merge 2 commits intomainfrom
tyler-griggs wants to merge 2 commits intomainfrom
Conversation
Adds a new policy loss function that wraps the PPO clipped surrogate
in an exponential tilt, focusing optimization on worst-case tokens.
L_dro = (1/beta) * log(E[exp(beta * L_ppo)])
Configurable via policy_loss_type: "dro" with dro.beta controlling
the degree of robustness (default 0.1).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add "sum" to reduce_loss (simple masked sum, no averaging) - Update ppo_policy_loss assertion, config validation, and docstring to include "sum" as a valid loss_reduction option - Recreate test file covering only functions that exist on this branch: DRO loss (5 tests), sum reduction (4 tests), zero-variance filter (4 tests), config fields (1 test) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
L_dro = (1/beta) * log(E[exp(beta * L_ppo)])policy_loss_type: "dro",dro.beta: 0.1Test plan
🤖 Generated with Claude Code