Different number of node inputs for SplineConv

[hkim716]

Hi Matt,

I have a question about SplineConv in mnist_graclus.py example.
For mnist superpixel dataset, the number of nodes are identical because each sample has 75 nodes.

My question is :
If there are mixed data samples that have different number of nodes (like 60~75 nodes) and edge_index, is it possible to use those samples as an input of SplineConv?

[rusty1s]

Yes, SplineConv (and any other GNN layer in PyG) is agnostic to the number of nodes and supports a varying number of nodes across graphs within a mini-batch.