Temporal Iterators cannot deal with unweighted networks (without edge attributes)?

[pauvilasoler]

Hello!

Besides the point raised in #279 that the temporal iterators are currently mapping the edge_weights as edge_attr, I guess there is no option (currently) for using unweighted networks when there are no other edge attributes (see below):

In the case where the (unweighted network) would have edge attributes, this exceptions would not be raised as (as is currently set up) these edge attributes would be passed as edge_weights (as mentioned in #279, edge_weights can hold multiple features per edge given the mapping to edge_attr of the underlying torch_geometric.data.Data > it would still work in the case with only one (non-weight) edge attribute as well)

However if the underlying temporal network contains neither weights (unweighted) nor edge attributes then check_temporal_consistency() will raise a TypeError.

Other than this, I also agree with #279 that maybe it would be best to rename edge_weights into edge_attr for better consistency with PyG.

Thanks a lot!