Open
Conversation
Introduces a new mechanism that adjusts per-token advantages based on KL divergence from a reference model. Tokens where the policy has drifted more get reduced advantages, while tokens that drifted less get increased advantages. The adjustment is zero-mean (centered) across tokens. New parameters on LocalBackend.train(): - kl_penalty_coef: coefficient for the adjustment (0.0 = disabled) - kl_penalty_reference_step: use a specific checkpoint step as reference - kl_ref_adapter_path: use an arbitrary LoRA adapter path as reference Also fixes a pre-existing bug in preprocessing/inputs.py where warmup config used incorrect field names (lr → learning_rate, kl_coef → kl_penalty_coef). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
LocalBackend.train()parameters:kl_penalty_coef,kl_penalty_reference_step, andkl_ref_adapter_pathpreprocessing/inputs.pywhere warmup config used incorrect field names (lr→learning_rate,kl_coef→kl_penalty_coef)Test plan
uv run prek run --all-files)kl_penalty_reference_step=0completed successfully on Kubernetes H200 GPUs, all 20 steps each🤖 Generated with Claude Code