Skip to content

fix: attach pipeline noise filter to image_processor logger#594

Merged
vortex-captain merged 1 commit into
mainfrom
reny/image_processor_warning
May 12, 2026
Merged

fix: attach pipeline noise filter to image_processor logger#594
vortex-captain merged 1 commit into
mainfrom
reny/image_processor_warning

Conversation

@vortex-captain

Copy link
Copy Markdown
Contributor

Fixes #420.

Summary

  • _PipelineNoiseFilter listed "Using a slow image processor" in its _SUPPRESSED tuple but was only attached to transformers.pipelines.base. The message is emitted by transformers.models.auto.image_processing_auto, so the suppression never fired.
  • Re-attach the filter to both loggers via a loop, restoring the behavior from 324407d.

The 'Using a slow image processor' suppression was silently broken because
_PipelineNoiseFilter was only attached to transformers.pipelines.base, but
the message is emitted by transformers.models.auto.image_processing_auto.

Re-attach the filter to both loggers so the suppression fires.
@vortex-captain vortex-captain requested a review from a team as a code owner May 12, 2026 05:47
@vortex-captain vortex-captain merged commit 5ff4018 into main May 12, 2026
9 checks passed
@vortex-captain vortex-captain deleted the reny/image_processor_warning branch May 12, 2026 05:54
ssss141414 pushed a commit that referenced this pull request May 15, 2026
Fixes #420.

## Summary

- `_PipelineNoiseFilter` listed `"Using a slow image processor"` in its
`_SUPPRESSED` tuple but was only attached to
`transformers.pipelines.base`. The message is emitted by
`transformers.models.auto.image_processing_auto`, so the suppression
never fired.
- Re-attach the filter to both loggers via a loop, restoring the
behavior from 324407d.

Co-authored-by: Yi Ren <reny@microsoft.com>
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.

Bug: 'Using a slow image processor' suppression silently broken in _warnings.py

3 participants