to_networkx() arguments require iterable of str

[jackychen08]

Currently, to_networkx() arguments such as node_attrs and edge_attrs require " iterable of str". This seems unintuitive as lots of features are stored as floats or ints in pytorch tensors. Would it make more sense to update this function to be able to accept tensors of floats/ints ?

[rusty1s]

This just refers to the names of attributes stored as part of Data, so I am not sure how one could support floats or ints here. Can you clarify?

[jackychen08]

From reading the documentation here , it looks like node_attrs, edge_attrs, etc should be passed in separately. Ie, the function call would look like this:
to_networkx(data, node_attrs = data.node_attrs, edge_attrs = data.edge_attrs). Do I not need to pass these attributes explicitly?

[rusty1s]

No, this looks fine. data.node_attrs and data.edge_attrs are of type List[str], so this matches with the function signature of to_networkx. Can you clarify your issue for me?