how to train model in multiple GPUs

[chenhao-user]

How to train this in multiple GPUs? I tried using multiple GPUs to train the model via DataParallel , but DataParallel will set self.net_recog.training to True，which causes an error to be reported when running the compute_losses_self。