how to find a reason for undirected graph for Data(...) format in PyG

[Amadeus-System]

Hi. I am a researcher in condensed matter physics.
I am preprocessing crystal graph dataset to Data format in PyG.

My Dataset should be converted to undirected, multi graph (which is possible to have multiple edges between nodes, but each edge between two nodes can have different edge_attr).

I have tested how Data() format in PyG work with simple graph example, after that I have created preprocessed custom dataset for my own.

However, there is some unknown reason.

as you can see in the above figure, each Data() format represents undirected, multi edge graph object. but when I use 'is_undirected()' method to check whether the graph is undirected, it returns false.
Since each Data() represents a single undirected-multi graph, it has repeated edge_index and repeated edge_attr.
And indexes of edge_index and edge_attr are corresponed, of course.

I have checked edge_index and edge_attr. but I cannot see any problems.
could you tell me how I can check the reason which make my dataset directed ??

[rusty1s]

Thanks for reporting. Our is_undirected functionality can be found here: https://github.com/pyg-team/pytorch_geometric/blob/master/torch_geometric/utils/undirected.py#L11-L51

Can you copy-paste the functionality and see where it might break on your end? Alternatively, you can provide a simple example in which is_undirected breaks for you. It might be related to using torch.all rather than torch.allclose.