我刚开始使用Pytorch，还在努力理解它的基本概念。
如果我在GPU上有一个网络n，它产生一个输出Tensorout，它可以直接打印到stdout吗？或者应该先将它移到cpu，或者在打印之前将它从图形中分离出来？
尝试了以下涉及.cpu()和.detach()的几种组合

import torch.nn as nn
import torch

class Net(nn.Module):
    def __init__(self):
        super().__init__()
        self.layers = nn.Sequential(
            nn.Linear(5, 10),
            nn.ReLU(),
            nn.Linear(10, 10),
            nn.ReLU(),
            nn.Linear(10, 3),
        )

    def forward(self, x):
        return self.layers(x)

device = torch.device("cuda:0")  # assume its available
x = torch.rand(10, 5).to(device)
net = Net().to(device)

# Pretend we are in a training loop iteration

out = net(x)
print(f"The output is {out.max()}")
print(f"The output is {out.max().detach()}")
print(f"The output is {out.max().cpu()}")
print(f"The output is {out.max().cpu().detach()}")

# continue training iteration and repeat more iterations in training loop

我得到了相同的输出为所有4种方法。哪种是正确的方式？