perf(rollout): six independent rollout-path fixes and knobs - #198
Open
Rockdu wants to merge 6 commits into
Open
Conversation
This was referenced Aug 21, 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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
Six independent rollout-path improvements, one commit each:
perf(rm_hub).remote()serialises the payload into plasma in the calling thread, so submitting on the event loop stalled every concurrent requestperf(rm_hub)fix(router)/remove_worker; engines call it on shutdown and it fell through the catch-all, proxied to a worker as a generate requestfix(fsdp)strict=Falseand discardedset_state_dict's result, so a checkpoint whose keys match nothing "loaded" silentlyperf(rollout)--rollout-request-stagger: one-shot jitter before slot acquisition; co-admitted requests stay in phase for the whole rollout (measured: median inter-arrival gap 0.00 s, 16 arrivals in the first 5% of each rollout)feat(rollout)--rollout-video-dtype {keep,uint8}: opt-in engine-side quantisation of the rollout video, default unchanged; needs the sglang counterpart to take effect and is harmless without itValidation
log_prob_mean_abs_diff3.0e-05 (in the verified band), 0 faults; with uint8 on,resp_bytes1384 -> 436 MiB andtrain_wait1010-1140 s -> 486-499 s.pre-commit runclean on all touched files;pytest tests/fast: 215 passed.Checklist
pre-commit run --all-filespasses — run on touched filespytest -xis green — tests/fast, 215 passedpython3 train.py --helpstill parses