Deriving The Structure Of A Pytorch Network

For my use case, I require to be able to take a pytorch module and interpret the sequence of layers in the module so that I can create a “connection” between the layers in some

Solution 1:

The information you are looking for is not stored in the nn.Module, but rather in the grad_fn attribute of the output tensor:

model = mymodel(channels)
pred = model(torch.rand((1, channels))
pred.grad_fn  # all the information is in the computation graph of the output tensor

It is not trivial to extract this information. You might want to look at torchviz package that draws a nice graph from the grad_fn information.