[Bugfix] harden Omni weight snapshot downloads#4139
Open
tzhouam wants to merge 2 commits into
Open
Conversation
Serialize full-repo Hugging Face snapshot materialization with a repo-wide node lock and retry incomplete snapshots so CI does not race on shared caches. Co-authored-by: Cursor <cursoragent@cursor.com>
|
Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits. |
tzhouam
changed the title
Collaborator
|
the ci fails due to qwen-image & HY-image acc, please check whether they are related |
Gaohan123
removed
nightly-test
Collaborator
|
please check the ci failures again an confirm whether it is not related |
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
3 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Follow-up to #4076. That PR hardened diffusion pipeline prefetching, but Buildkite still shows Qwen3-Omni online serving failing during stage-0 startup because the generic Omni HF snapshot path can still expose a half-materialized shared cache to transformers.
download_weights_from_hf_specific()usingfcntl.flockwith an atomic dotfile-lock fallback.Failure log: https://buildkite.com/vllm/vllm-omni/builds/11013/canvas?jid=019e90e1-a221-4fac-9049-9ac2492801ed&tab=output
Test plan
python -m py_compile vllm_omni/model_executor/model_loader/weight_utils.pypython -m ruff check vllm_omni/model_executor/model_loader/weight_utils.pypython -m ruff format --check vllm_omni/model_executor/model_loader/weight_utils.pyMade with Cursor