How to obtain link prediction for a single node in the bipartite_sage_unsup.py example?

[tirzak]

Hello,
I am a beginner; I am having a bit of trouble understanding the bipartite_sage_unsup.py example. How would someone obtain link prediction for a single user node after the model has been trained?

Thank you for your time, and thank you for making such a useful library.

[rusty1s]

What you can do is run your GNN as normal, and define the edges you are trying to predict as part of edge_label_index. So if you want to make predictions for a single user but all movies, that would be

row = torch.tensor([user_i] * num_movies)
col = torch.arange(num_movies)
edge_label_index = torch.stack([row, col], dim=0)

[tirzak]

Thank you very much