Questions regarding graphs with multiple edges

[yestinchen]

Hi,

Thanks for devloping such a great platform for GNN.

I would like to use GNN to learn from a graph with multiple edges and I haven't found any discussions from the document regarding multiple edges between two nodes.
I am looking at the implementation of GeneralConv Layer, there are some questions I would like to confirm:

1. It seems that to support multiple edges, we only need to add duplicate entries in edge_index, and add corresponding edge ettributes in the edge_attr?
2. Can we do the same thing for graphs with self-loops? Is it okay if say I have one node with multiple self-loop edges but with different features? Assuming all edges are of the same type.

Thank you!

[rusty1s]

Yes, your understanding is absolutely correct.

[yestinchen]

Thanks for the quick response!