Converting networkx graph to PyG Data issue.

🐛 Bug

Converting nx.DiGraph to Data raises ValueError: only one element tensors can be converted to Python scalars

     98     for key, item in data.items():
---> 99         data[key] = torch.tensor(item)
    100 
    101     data['edge_index'] = edge_index

To Reproduce

Steps to reproduce the behavior:

1. Convert PyG Data to a networkx Graph using pytorch_geometric.utils.to_networkx providing edge attributes as a parameter.
2. Try to convert it back to PyG Data. from_networkx(to_networkx(pyg_graph, node_attrs=['pos', 'x'], edge_attrs=['edge_attr']))

Expected behavior

I would expect it to work properly. Also there is no possibility now to convert graph features to nx.Graph and back.

Environment

• OS: Ubuntu 18.04.1 LTS
• Python version: 3.6.9
• PyTorch version: 1.1.0
• CUDA/cuDNN version: 10.0.130
• GCC version: 7.4.0
• Any other relevant information: N/A

Additional context

N/A

[rusty1s]

Fixed in master! Sorry 😃

Edit: Steps to reproduce:

import torch
from torch_geometric.data import Data
from torch_geometric.utils import erdos_renyi_graph, to_networkx, from_networkx

edge_index = erdos_renyi_graph(50, edge_prob=0.2)
x = torch.randn(50, 64)
edge_attr = torch.randn(edge_index.size(1), 32)
data = Data(edge_index=edge_index, x=x, edge_attr=edge_attr)

print(data)
G = to_networkx(data, node_attrs=['x'], edge_attrs=['edge_attr'])
data = from_networkx(G)
print(data)

[kunal-bhadra]

I have converted it into a homogenous Networkx graph and now all nodes/edges are in integers. However, I can’t find any documentation on the Networkx homogenous -> heterogenous conversion. Do we have to implement this on our own with the Networkx modules from scratch?