torch.save(predictions, predictions_path, pickle_module=dill) system dead

Question

I trained the whole network and got the parameters. When I use one pth to infer the result on a test dataset, just testing, by debugging, each time when it goes to "tor