Why are Pytorch models invoked as functions?
For a long time, I have wondered why we invoke a Pytorch model itself, like, model(input,targets)
instead of invoking its forward(inputs, targets)
method. Today I went looking for answers to this design oddity.