When training yolov8, the device specifies multi-GPU training and reports an error (single card can run):
yolo task=detect mode=train model=yolov8n.yaml device=0,1,2,3
Reason: pytorch version problem, in order to adapt to the edge computing box, the 1.7 version of pytorch is used
Solution: Upgrade the pytorch version, probably after 1.9, the official recommendation is torch>=2.0.0 version, pytorch upgrade command can be found in Previous PyTorch Versions | PyTorch