GAE graph autoencoder

[thegodone]

Can you implement the graph basic autoencoder that needs to do InnerProduct of the latent space to reconstruct Adjacency matrix ?
https://pytorch-geometric.readthedocs.io/en/latest/generated/torch_geometric.nn.models.GAE.html
basically the encoder part can be GIN, GAT etc... and we need the InnerProduct finish the look. Of Course a loss need to be added for the graph to reconstruction (X,A) input / output by similarity.