Training a GAT for Graph classification

[15235007059]

Hello, author. I have 969 samples, each of which corresponds to the same graph structure and has 6888 nodes. Want to realize graph classification on the basis of your source code, but tried for a long time or did not find a suitable way to deal with data. I don't know how to convert raw data from.csv format to GAT supported such as feat/label/graph_id. npy / graph.json. Your help will be greatly appreciated!

[qycss]

I didn't find a good way to convert .csv to .csr, so I rewrite a function to deal with my data and transfer it to np.matrix before return.