Lesson 5: forward() method signature

Part 1 (2018)
1

Hi,

As per my understanding forward is an special function which gets called by PyTorch to do forward pass calculations. I ma confused regarding the signature of this method.

In lesson 5 notebook we see two different implementations of the method.:

    class DotProduct(nn.Module):
        def forward(self, u, m): return (u*m).sum(1)

    class EmbeddingDot(nn.Module):
        ...
        ...
        def forward(self, cats, conts):
            users,movies = cats[:,0],cats[:,1]
            u,m = self.u(users),self.m(movies)
            return (u*m).sum(1).view(-1, 1)

I have the following doubts:

  1. In the second implementation what is meant by arguments cats and conts?
  2. And how are they passed to the forward method?
  3. Why cant we directly pass users and movies vectors as in the first implementation?

Thanks.