I am experimenting with GNN and am sort of stuck at creating a dataset/loader with torch_geometric for training in mini batches. Domain is molecules/drugs…
Here is where am at…
- I have created a python list of
a. Nodes(a list of arrays of various shapes) - train_N … eg…(shapes.).[(42,10),(42,10),(65,10)…]
b. and a list of corresponding adjacent matrices - train_adj…eg…[(42,42),(42,42)…(65,65)…]
c. y values shape (200,5000)
How do I convert all these into a torch_geometric DataLoader and use it in my model training in batches ?
How do I