Multistream network with lateral connection in pytorch

I need to work on a project for which I need two different branches for a neural network using PyTorch, fusing the output at a late stage. I would like to implement lateral connections so that the networks are able to share information, but I couldn’t find any documentation about it. Is it possible to implement these types of connections in PyTorch in general? And if the two branches are a Convolution branch and a GNN/GCN branch?