chore: finish the legacy-ReFL deletion sweep - #293
Merged
Conversation
Post-merge sweep completing #281's blast radius (these fixes raced the squash-merge and landed on the closed branch): - experimental/refl README: list the deliberate capability cuts vs the deleted core path (periodic eval + eval_rewards suites, reward_fraction placement, debug knobs) with the bring-back rule, so reviewers don't rediscover them as migration gaps. - Rewrite docstrings that still referenced the deleted RewardBackpropTrainer / ReFLPolicy / draft_generate (experimental trainer/roles/example yaml, sd3 build_conditions, eval_suites). - CODEOWNERS: drop the two rules covering deleted paths. grep-level zero references to the deleted symbols remain.
haonan3
requested review from
CjhHa1,
Ideny42,
celve and
leviking98z-rgb
as code owners
August 2, 2026 03:25
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Post-merge sweep completing #281's deletion blast radius — these fixes raced the squash-merge and landed on the already-closed branch, so main never got them:
eval_rewardssuites from feat(trainer): add periodic eval to PE / ReFL / UnifiedModel #202,reward_fractioncross-slab placement,param_checksum/activation_checkpoint_vaeknobs) with the bring-back rule — so reviewers stop rediscovering them as migration gaps (one external review already misread the rewrite as "fork drift").RewardBackpropTrainer/ReFLPolicy/draft_generate(experimentaltrainer.py/roles.py/ wan21 example yaml,unirl/models/sd3/pipeline.pybuild_conditions,unirl/trainer/eval_suites.py×2)./unirl/train/refl/,/unirl/train_refl.py).No behavior change — docs, comments, and review-routing rules only.
Related Issue
Follow-up to #281 (same genre as the #290 post-merge sweep).
Test Plan
grep -rn 'train_refl\|RewardBackpropTrainer\|ReFLPolicy\|draft_generate\|models\.draft' --include='*.py' --include='*.yaml' --include='*.md' .→ zero references to deleted symbols outside historical notes.grep -n refl .github/CODEOWNERS→ empty.pre-commit run --files <touched files>→ pass.Compatibility / Risk
None — no executable code changed. AI-assisted (Claude); overlap checked against #290's refl nits (disjoint files).
🤖 Generated with Claude Code