Open
Conversation
vivekkalyan
approved these changes
Mar 17, 2026
Collaborator
vivekkalyan
left a comment
vivekkalyan
left a comment
There was a problem hiding this comment.
thanks for catching this!
angkywilliam
approved these changes
Mar 17, 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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
It fixes a client-side W&B run leakage bug in ART.
Before the change in
src/art/model.py, ART would callwandb.init(...)without telling W&B to create a fresh run. In a single Python process, if one ART model had already opened a run, W&B could return that existing active run for the next model. The result was that metrics from model B, model C, and so on could be logged into model A’s W&B run.That matters most for the serverless case: one client process can create and manage multiple training jobs, potentially across multiple GPUs. Those jobs need separate W&B runs. The fix makes ART open a distinct W&B run per model with
reinit="create_new"and defines metrics on that specific run object instead of module-global W&B state. So metrics stay attached to the correct run name instead of being silently merged into the first one.