How to carry out the batch-wise training while using layer defined in DGL?

Things in the last sentence I want to express are that:
(1) the graph embedding is learned for the down-stream regression problem;
(2) fail to know how many classes exist among graphs, but tends to be many i.e. large value of “num_class”;
(3) could I set the “num_class” attribute larger to prevent information?
(4) or only set the “him_dim” larger as you recommended?

Then how do you want to train the model if the number of classes/task is also different?

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