Your
python polygon_train.py --weights polygon-yolov5s-ucas.pt --cfg polygon_yolov5s_ucas.yaml --data polygon_ucas.yaml --hyp hyp.ucas.yaml --img-size 1024 --epochs 3 --batch-size 12 --noautoanchor --polygon --cache
Or my
python polygon_train.py --weights polygon-yolov5s-ucas.pt --cfg polygon_yolov5s_ucas.yaml --data polygon_dir4.yaml --hyp hyp.ucas.yaml --img-size 640 --epochs 3 --batch-size 100 -noautoanchor --polygon --cache
Its progress is always 0%, and there are no error messages. My GPU is 4090. I configured all the information according to your instructions and also installed the libraries required for your cuda using cd utils/iou_cuda && python setup.py install. Did I miss anything else?
It seems that the current training code does not support multi-GPU parallel training, and the training can be carried out normally when the single self switch is a single card
Your
python polygon_train.py --weights polygon-yolov5s-ucas.pt --cfg polygon_yolov5s_ucas.yaml --data polygon_ucas.yaml --hyp hyp.ucas.yaml --img-size 1024 --epochs 3 --batch-size 12 --noautoanchor --polygon --cache
Or mypython polygon_train.py --weights polygon-yolov5s-ucas.pt --cfg polygon_yolov5s_ucas.yaml --data polygon_dir4.yaml --hyp hyp.ucas.yaml --img-size 640 --epochs 3 --batch-size 100 -noautoanchor --polygon --cache
Its progress is always 0%, and there are no error messages. My GPU is 4090. I configured all the information according to your instructions and also installed the libraries required for your cuda usingcd utils/iou_cuda && python setup.py install
. Did I miss anything else?