feat: extend GRPO with ProRL/DAPO controls and document RL/inference stack #228
Conversation
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
…stack Enhance GRPOTrainer with advanced RL features: - Add asymmetric clipping, per-token weighting, length shaping - Support multiple advantage estimators (mean, median, GAE) - Implement KL/entropy guards and reference model resets - Add group-normalized advantages and DAPO length shaping Improve DPO metrics logging: - Add KL divergence, entropy, and logit margin tracking - Ensure consistent dashboard metrics across trainers Documentation updates: - Add RL_ALGORITHMS.md documenting all RL algorithms - Add INFERENCE_ENGINES.md documenting vInference/vSurge/vWhisper - Add RL_UPDATES.md with release notes - Add VINEPPO_IMPLEMENTATION_PLAN.md for future work - Update docs/trainers/grpo.md with new configuration options Testing: - Add tests/trainers/grpo_utils_test.py for utility functions
Author
|
lets see how codex did |
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Outdated suggestions cannot be applied.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
GRPOTrainerto expose ProRL/DAPO-style controls (asymmetric clipping, per-token weighting, length shaping, KL/entropy guards, selectable advantage estimators)DPOTrainermetrics with KL/entropy/logit margin for consistent dashboardsRL_ALGORITHMS.md), inference engines (INFERENCE_ENGINES.md), release notes (RL_UPDATES.md), and provide a VinePPO roadmap (VINEPPO_IMPLEMENTATION_PLAN.md)Changes
easydel/trainers/group_relative_policy_optimization/grpo_config.py: New config knobs (advantage estimator, clip bounds, length shaping, KL resets, sampling safeguards)easydel/trainers/group_relative_policy_optimization/grpo_trainer.py: Preprocess rewards with length shaping, compute advanced metrics, manage reference resets (ProRL-style)easydel/trainers/group_relative_policy_optimization/_fn.py: PPO-style ratio loss with asymmetric clipping and token weightingeasydel/trainers/training_utils.py: Shared helpers for group advantages, length shaping, EMA trackingeasydel/trainers/direct_preference_optimization_trainer/_fn.py: Richer logging metricstests/trainers/grpo_utils_test.py: Unit coverage for new helpersdocs/trainers/grpo.md,RL_ALGORITHMS.md,INFERENCE_ENGINES.md,RL_UPDATES.md,RL.md,VINEPPO_IMPLEMENTATION_PLAN.mdTesting
Follow-ups
VINEPPO_IMPLEMENTATION_PLAN.md)Notes
ejkerneloptimisations