Conversation
Arist12
requested review from
Shi-Dong,
Zhichenzzz,
fzyzcjy,
maocheng23,
yueming-yuan and
yushengsu-thu
as code owners
August 22, 2026 00:15
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
A resume passed --lora-adapter-path straight to SGLang as lora_paths, but that directory holds miles native shards plus a best-effort PEFT export. When the export was skipped or written in a layout the serving loader does not expect, the engine failed to start. Start every engine from the base model instead. train.py and train_async.py both publish the trainer adapter unconditionally before the first rollout, so the adapter arrives over the normal weight-sync path and native-only and PEFT-producing resumes take the same route. That publish now upserts rather than unload-then-load, so an adapter is never briefly absent from a live engine. Upsert is only wired up on the from_distributed route, so the colocated from_tensors path keeps its explicit unload. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Arist12
force-pushed
the
fix/lora-disaggregated-resume-adapter
branch
from
September 1, 2026 20:30
f7fe762 to
7665877
Compare
Contributor
Author
|
Closing against the refactor. The |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Part of #2705.
Problem
A disaggregated LoRA run resumed with
--lora-adapter-pathcan fail before its first rollout: native-only checkpoints cannot be preloaded by SGLang, partial best-effort PEFT output can look preloadable, and a preloaded adapter collides with the mandatory initial publish.Change
Normal train drivers start resumed rollout engines from the base model and let the mandatory initial weight publish install the trainer's restored adapter. Publish single LoRA adapters with SGLang's existing
upsert=Truecontract so fresh startup and later refreshes use one path without unload churn.--debug-rollout-only, which skips the publish, remains a blocker before this PR is ready.Validation
384tensors each), optimizer, scheduler and global-dataset cursor fromiter_29; continued at rollout 30, completed two optimizer steps, savediter_30, and republished the trained adapter (grad_norm=0.0152, train/rollout abs diff0.0210-0.0237)iter_4anditer_14, restored288tensors plus optimizer/scheduler/data cursor, and completed throughiter_290.711 → 0.789, truncation0.211 → 0.078; train/rollout abs diff mean0.01265, max0.01419