Explainer ModelConfig - Different task levels

[acperson]

Hi,

I am new to pytorch geometric and have a question about the task_level parameter in ModelConfig for the explainer.
In particular, I was wondering what the difference between node and edge task_levels are i.e. where they are called within the Explanainer Algorithm and how they influence the explanation.
I was looking through the source code but could not find where the task_level parameter is inferred.

Thank you so much for your help!

[rusty1s]

In most cases, the PyG explainers are agnostic to the underlying task, and this argument is just not used. For example, GNNExplainer does not require this argument at all. In other explainers, we use TaskLevel.node for example to reduce the the graph to k-hop subgraphs around nodes to explain.