Repository navigation
Keep MM validation sample counts intact - #3200
dajiaohuang wants to merge 1 commit into
Conversation
|
Warning Review limit reachedYou've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Next included review available in 12 minutes. View limit detailsLimit details: You’ve used all 8 included reviews currently available. Review configuration: ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Advanced Run ID: 📒 Files selected for processing (3)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
This changes behavior in a maintenance-only integration (MM/Labelbox/COCO) rather than fixing a defect, and it has no owner review or integration run behind it. Holding it: it needs the integration owner to decide whether the new behavior is wanted. No action needed from you now. |
Fixes #3199.
MMDet/MMSeg evaluator wrappers count samples, but their validation
__len__methods divided the sample count by batch size and returned the result as a sample count. The actual validation DataLoader uses the full underlying dataset whendrop_last=False, so evaluators could load only part of the ground truth and omit predictions. Return the full validation dataset length; drop behavior stays with the DataLoader configuration.Validation:
__len__methods from both source files with batch_size=2.git diff --check: passed.No full MMDet/MMSeg runtime is configured on this host.