Multiple forward() calls in one step should be separate in the interface

[spfrommer]

This is pretty tricky to implement so probably not a priority for now. In the scenario where the root module's forward method is invoked multiple times before the loss backward, right now the resulting i/o histograms are superimposed. Really, you want to tease these apart somehow in the interface. Not sure what's the right implementation: probably a drop-down list...