fix: preserve per-sample training metadata across DP transport - #2403
Open
heyxingxing wants to merge 1 commit into
Open
heyxingxing wants to merge 1 commit into
heyxingxing wants to merge 1 commit into
Conversation
This branch has not been deployed
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Per-sample training metadata disappears before reaching the trainer: conversion ignores it when only later samples provide it, and DP transport omits the field even when conversion populated it. Include metadata when any sample provides it, retain
Noneentries, and partition it alongside the other per-sample fields.Closes #2400
Validation:
PYTHONPATH=. python tests/test_process_rollout_data.py— 20 passed. Twelve added regression cases cover DP sizes 1/2, static/dynamic batching, and complete/mixed/absent metadata; eight of them fail on the base implementation with missingmetadata. Tests exercise the real conversion, scheduler and consumer with serving/telemetry imports and Ray object-store calls isolated. The existing CPU CI registration is reused.Ruff 0.14.7, Black 24.3.0, isort 5.13.2 and
git diff --checkpassed. CPU validation only (Python 3.12.13, macOS arm64); no GPU or multi-process Ray run.