how to use pyG GNN explainer to explain graph level prediction and visualize the graph.

[ravi5233]

I trained the TransformerConv model, i need to check prediction of model with graph level. My dataset is molecular structures and i want to visualize the atom weight heatmap which part of molecular structure responsible for a particular predicted instance. I would like to seek help from pyG community. Thank you for your help.

[rusty1s]

Did you take a look at the return_attention_weights argument of TransformerConv?