Why fc_coord is Linear() Instead of nn.linear in Regression_nf_3d.py?

[2015wan]

1. Why design a separate linear layer for the output of the coordinate mean, and what is the difference between it and the ordinary fully connected layer?
3. Why is this happening?
   x_norm = torch.norm(x, dim=1, keepdim=True)
   y = y / x_norm

[2015wan]

In this class RegressFlow3D(nn.Module) Why when testing
pred_jts[:, :, 2] = pred_jts[:, :, 2] - pred_jts[:, self.root_idx:self.root_idx + 1, 2]
Why do you have to subtract like this