PyTorch's convolutional layers do no accept asymmetric padding, but we can do it with F.pad, ... Similar processing can be achieved with the modules nn.
確定! 回上一頁