How to pass the updated edge_attr to the next layer?

[Bowen-n]

It seems that self.aggregate only returns the updated node features.

I want to update my edge_attr and pass is to the next layer. How to implement my MessagePassing model ?

[Bowen-n]

For example, how to implement GatedGCN-E mentioned in paper Benchmarking Graph Neural Networks.

[rusty1s]

If you want to also update your edge features and return it as part of your GNN layer, take a look at edge_updater functionality within MessagePassing, see here. Here is an example on how to use.