Creating Custom Graph - Confused with RGCN Tutorial

Hello DGL Community!

I am new to DGL and am following the RGCN tutorial here: https://docs.dgl.ai/en/0.4.x/tutorials/models/1_gnn/4_rgcn.html

The tutorial walks through the AIFB dataset, and I tried modifying the tutorial notebook to work with my own dataset. However, I am confused as to what the parameters for the model mean and am having a hard time figuring out what they are (e.g. num_rels, num_classes, labels).

I assumed that num_rels was the number of unique edge types, but when I look at the AIFB unique edge types versus data.num_rels, the numbers do not match. Also, I’m guessing that num_classes is the number of node classes?

My Jupyter notebook is in the link below and is completed to the best of my guessing, which runs, but it sigkills on my end without any error messages. I made the train, validation, and test batches very small (10, 5, 5) so I’m not sure why this is happening!

Help would be much appreciated!

Sorry for the unpleasant experience.

  • num_rels is the number of relations. It seems that Model takes data.num_rels for the num_rels argument here?
  • num_classes is the number of node classes as you said.
  • Can you also upload ‘network.csv’?

Do you have any suggestions for improving the tutorial? E.g. will it be better to have doc for the parameters in the tutorial?

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.