How to create a 'multiply layer' which supports backprop?

4

If I use torch.multiply will the backpropagation be handled correctly? I am currently using it to multiply two variables (output of convolutions) inside the forward method of a custom module (extends nn.Module).