有大佬知道cuda:0 device type tensor to numpy怎么解决

[xiaobai1feike]

can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.

[zhunzhun123]

What exactly is the error

[MarkLi1214]

在tensor变量后面加.cpu()

[jibrilhp]

i changed it in lines 120-121 :

 # Compute loss at each stage
        loss_char =  torch.sum(torch.stack([criterion_char(restored[j], target) for j in range(len(restored))]))
        loss_edge = torch.sum(torch.stack([criterion_edge(restored[j], target) for j in range(len(restored))]))
        loss = (loss_char) + (0.05*loss_edge)

and it's worked

[6jybuchiyu]

i changed it in lines 120-121 :

 # Compute loss at each stage
        loss_char =  torch.sum(torch.stack([criterion_char(restored[j], target) for j in range(len(restored))]))
        loss_edge = torch.sum(torch.stack([criterion_edge(restored[j], target) for j in range(len(restored))]))
        loss = (loss_char) + (0.05*loss_edge)

and it's worked

It works！ Thank you！

[Feecuin]

请问下改了这里之后你们训练会遇到损失降不下来的情况吗？我PSNR一直上不去，后面训练的都不如前面的很低很低，这是什么情况呢？