Using Nodes Features in Heterogeneus R-GCN

Hello everyone,

I am new to using DGL, and I must say I find its commands very user-friendly.

I’m currently working on a project using the EntityClassify and EntityClassify_HeteroAPI models (dgl/examples/pytorch/rgcn-hetero/model.py at master · dmlc/dgl · GitHub), but I am encountering difficulties utilizing the node features in my graph.

The model always trains on embeddings initialized with Xavier initialization, and I cannot figure out how to make it work with the actual node features in my graph.

My graph is a heterogeneous graph with nodes having features of different dimensions (though this shouldn’t be the issue, as I can’t make it work even when assigning a feature for each node type).

I apologize for the basic question, but as I mentioned, I’m just getting started with this.