Replace kl_penalty_reference_step with kl_penalty_step_lag#625
Open
angkywilliam wants to merge 13 commits intofeat/pipeline-klfrom
Open
Replace kl_penalty_reference_step with kl_penalty_step_lag#625angkywilliam wants to merge 13 commits intofeat/pipeline-klfrom
angkywilliam wants to merge 13 commits intofeat/pipeline-klfrom
Conversation
…ipelineTrainer - Rename parameter from `kl_penalty_reference_step` to `kl_penalty_step_lag` - `None` (default): uses step 0 as KL reference (anchor to initial model) - `>= 1`: uses `max(0, current_step - lag)` as reference (rolling anchor) - Add validation that kl_penalty_step_lag must be >= 1 if specified - Update existing tests and add new tests for lag computation 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add tinker variant of KL-penalized advantage training script and align model naming conventions (backend-random-coef) across both. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
vivekkalyan
force-pushed
the
feat/pipeline-kl
branch
from
5ef9eab to
aa3e6f7
Compare
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
kl_penalty_reference_steptokl_penalty_step_lagin PipelineTrainerNone(default): uses step 0 as KL reference (anchor to initial model)>= 1: usesmax(0, current_step - lag)as reference (rolling anchor)kl_penalty_step_lagmust be >= 1 if specifiedTest plan
test_pipeline_trainer_local_backend.py)🤖 Generated with Claude Code