Metrics are not displayed when training on the YOLO dataset

Metrics are not displayed when training on the YOLO dataset. I use the standard YOLO dataset, at least some metrics are shown during training
Epoch gpu_mem box obj cls pixl area dist total labels img_size
0/49 9.43G 0.124 0.02577 0 0.5841 0 0 0.2666 38 1024: 91%|████████▏| 49/54 [02:57<00:08, 1.67s/it]
And at vadidation everything is zero, although images and labels are loaded correctly.
Class Images Labels P R mAP@.5 mAP@.5:.95 BPR Occupy: 100%|█████████████| 7/7 [00:00<00:00, 14.39it/s]
all 50 0 0 0 0 0 nan 1

val: Scanning 'VisDrone/VisDrone2019-DET-val/labels.cache' images and labels... 50 found, 0 missing, 0 empty, 0 corrupted: 100%|██████| 50/50 [00:00

Reproduce the report with the standard YOLO dataset and the VisDrone validation data shown in the logs. Trace the training and validation entry points to determine why validation reports zero labels and zero metrics despite loading images and label files. Done means validation displays meaningful metrics for a correctly formatted dataset.

Written by the indexing model from the issue text.