Skip to content

docs: add release notes for vLLM race condition fix (PR #1590) - #1696

Merged
lbliii merged 1 commit into
NVIDIA-NeMo:26.04-stagingfrom
lbliii:lbliii/docs-aegis-guardrails
Apr 1, 2026
Merged

docs: add release notes for vLLM race condition fix (PR #1590)#1696
lbliii merged 1 commit into
NVIDIA-NeMo:26.04-stagingfrom
lbliii:lbliii/docs-aegis-guardrails

Conversation

@lbliii

@lbliii lbliii commented Mar 31, 2026

Copy link
Copy Markdown
Contributor

Description

Adds a bug fix entry to the 26.04 release notes for PR #1590, which fixed a race condition in CaptionGenerationStage and CaptionEnhancementStage where multiple workers initializing vLLM would race on the shared torch.compile cache directory. No user-facing API changes; documentation-only.

Usage

N/A — documentation-only change.

Checklist

  • I am familiar with the Contributing Guide.
  • New or Existing tests cover these changes.
  • The documentation is up to date with these changes.

@lbliii
lbliii requested a review from a team as a code owner March 31, 2026 18:53
@lbliii
lbliii requested review from VibhuJawa and removed request for a team March 31, 2026 18:53
@lbliii lbliii self-assigned this Mar 31, 2026
@greptile-apps

greptile-apps Bot commented Mar 31, 2026

Copy link
Copy Markdown
Contributor

Greptile Summary

This is a documentation-only PR that adds a bug fix entry to the 26.04 release notes for PR #1590, which resolved a race condition in CaptionGenerationStage and CaptionEnhancementStage.

  • Adds a new ### Video vLLM Setup Race Condition (PR #1590) section under the existing ## Bug Fixes heading, consistent with the structure of adjacent entries.
  • The description accurately captures the root cause (multiple workers racing on the shared torch.compile cache directory), the symptom (FileNotFoundError), and the fix (moving model initialization to setup_on_node() instead of setup()).
  • No API, code, or configuration changes are included; this is purely additive to the release notes.

Confidence Score: 5/5

Safe to merge — documentation-only change with no code impact.

The change is a single additive paragraph in a release notes file. The content is technically accurate, well-structured, and consistent with adjacent entries. No logic, API, or behavior is altered.

No files require special attention.

Important Files Changed

Filename Overview
fern/versions/v26.04/pages/about/release-notes/index.mdx Adds a new "Video vLLM Setup Race Condition (PR #1590)" entry under the Bug Fixes section; content is accurate, well-placed, and consistent with adjacent entries.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Multi-worker vLLM startup] --> B{Before fix}
    A --> C{After fix - PR #1590}

    B --> D["Each worker calls setup()"]
    D --> E["All workers race to initialize vLLM"]
    E --> F["Race on shared torch.compile cache dir"]
    F --> G["FileNotFoundError 💥"]

    C --> H["setup_on_node() runs once per node"]
    H --> I["Single initialization per node"]
    I --> J["Workers proceed safely ✅"]

    style G fill:#ff6b6b,color:#fff
    style J fill:#51cf66,color:#fff
Loading

Reviews (2): Last reviewed commit: "docs: add release notes for vLLM race co..." | Re-trigger Greptile

@VibhuJawa
VibhuJawa requested a review from weijiac0619 March 31, 2026 19:05
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants