Dividing a data object into batches

[atul1234anand]

Is there a way to divide a data object in pytorch geometric into batches without sampling?

[rusty1s]

Do you mean into a set of subgraphs? Take a look at NeighborLoader.

[atul1234anand]

NeighbourLoader samples some nodes per neighbourhood right? I dont want that. I would like to sample a bunch of nodes and the induced graph of those nodes.

[rusty1s]

If you don't want to sample a neighborhood, you can also do

nodes = torch.randint(0, data.num_nodes, (batch_size, ))
batch = data.subgraph(nodes)