Custom nn.Conv2d

autograd
3

Can i do it like this https://discuss.pytorch.org/t/custom-a-new-convolution-layer-in-cnn/43682 ?