Losses in Fasterrcnn end up becoming NAN during training.Why the model returning nan as output's Loss?

I replaced L1-smooth Loss in bounding box refinement state with IoU Loss and GIoU Loss in Fasterrcnn,but the result of class_loss ,regression_loss,rpn_class_loss and rpn_bbox_regression_loss is Nan.Thus,Fasterrcnn is stopping train. I use the same parameters as L1-smooth Loss.

debug is Nan

This my result of debug!