HeteroConv for empty node types

[strukturen]

Hi,

I am using HeteroConv with e.g. HeteroConv({('type1','to','type2'): some SAGEConv layer, ('type3','to','type2'): some SAGEConv layer}, aggr = 'sum') but not all my data samples have 'type1' nodes. Hence, currently if one sample misses a node type (e.g. 'type1') it cannot execute the forward function and throws me an error:
ValueError: Encountered invalid 'dim_size' (got '3' but expected >= '4')
The error does not appear if each data sample has 'type1' nodes.

What can I do to solve this respectively that it just ignores executing this layer for the specific dataset? My torch-geometric version is 2.6.1.

Thank you!