为什么训练完之后总会出现2023/09/28 04:07:35 - mmengine - ERROR - /home/jupyter/miniconda3/envs/mmyolo/lib/python3.8/site-packages/mmdet/evaluation/metrics/coco_metric.py - compute_metrics - 461 - The testing results of the whole dataset is empty.

[chenhan1hao]

为什么训练完之后总会出现2023/09/28 04:07:35 - mmengine - ERROR - /home/jupyter/miniconda3/envs/mmyolo/lib/python3.8/site-packages/mmdet/evaluation/metrics/coco_metric.py - compute_metrics - 461 - The testing results of the whole dataset is empty.

[Z-Kamito]

请问解决了吗？我也遇到这样的情况，原因是我改动了resnet，增加了一层输出层 ，除了改动resnet还有哪个地方需要改动吗

[Le0v1n]

1. Make sure to add meta_info to both train_dataloader and val_dataloader.
2. Ensure that model.bbox_head.head_module.num_classes is set correctly.

[wsy-yjys]

@Le0v1n What is meta info , how to add it, and is there any documentation?