Question about how to align dataset

[nickspark]

I have a question about your pretraining model, I see in your code, each cell in a batch is a molecule graph, however, in some GNN models, the trainable weights W's dimension depends on the size of the input graph, obviously, each molecule graph has different nodes, so I wonder, how do you align your pretrained model's trainable weights to different molecule with different node number in the downstream task.