Open
Conversation
Remove the Schulman KL estimator (beta * KL) that was added directly to the training loss. The kl_penalty_coef mechanism (advantage adjustment) remains as the preferred approach for KL regularization. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
corbt
approved these changes
Mar 9, 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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
betaparameter and Schulman KL divergence estimator (exp(r-n) - (r-n) - 1) that was added directly to the training losskl_penalty_coefmechanism (zero-mean advantage adjustment) remains as the preferred approach for KL regularizationChanges
src/art/types.py: Removebetafield fromTrainConfigsrc/art/loss.py: Removemean_klfromLossclass and the KL divergence computationsrc/art/local/backend.py: Removebetaparameter fromLocalBackend.train()src/art/serverless/backend.py: Removebetaparameter fromServerlessBackend.train()src/art/unsloth/train.py: Removebeta * mean_klloss addition andkl_divmetric loggingsrc/art/megatron/train.py: Removebeta * mean_klloss additionsrc/art/preprocessing/inputs.py: Removebetafrom warmup config overrideTest plan
uv run prek run --all-filespasses locally (ruff, ruff format, ty)test_backend_train_api.pypassed on H200 GPU cluster — model registration, trajectory gathering, training, and logging all succeeded🤖 Generated with Claude Code