Closed ver228 closed 1 year ago
That's because GCNConv
does not support heterogenerous graphs, check here for all supported operators (marked as bipartite
).
Thanks a lot! Now this make sense :)
Do you know if there is a way to access to this information programatically? I am trying to dynamically change the base GNN for GCNConv, and it will be nice to throw a more informative error if somebody tries to do the same mistake i was doing.
Actually, the error message has been added in PyG via #7637, you might need to upgrade your PyG verison :)
Thanks! I see the change hasn't come to a release version yet since the last release version is from the 27 Apr while the change was merged on the 23 Jun. I am happy to wait to the next release. Thanks for your help!
🐛 Describe the bug
It seems that if only GCNConv are used on a HeteroConv the forward pass fails.
Below is a minimum example:
The error message I have is:
Thanks for your help!
Environment
conda
,pip
, source): piptorch-scatter
): NA