question about torch_scatter

[Byun-jinyoung]

Dear pyg users,
I've generated graph by using pytorch geometric Data for regression problem.
Below is an example of the shape of data
Data(x=[6900, 40], edge_index=[2, 24932], edge_attr=[24932], y=[1], pos=[6900, 3])

However, when I tried to train my model, I got the error. Thus I input only one data into the model so that I got the error as below.

File "test.py", line 206, in <module>
    main()
  File "test.py", line 149, in main
    model(data)
  File "/home/byun/anaconda3/envs/pyg_torch/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
    return forward_call(*input, **kwargs)
  File "/home/byun/PROject/graphDPI/GADGET/model/pyg_model.py", line 101, in forward
    x = layer(x, edge_index)
  File "/home/byun/anaconda3/envs/pyg_torch/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
    return forward_call(*input, **kwargs)
  File "/home/byun/anaconda3/envs/pyg_torch/lib/python3.7/site-packages/torch_geometric/nn/conv/gcn_conv.py", line 165, in forward
    self.improved, self.add_self_loops)
  File "/home/byun/anaconda3/envs/pyg_torch/lib/python3.7/site-packages/torch_geometric/nn/conv/gcn_conv.py", line 62, in gcn_norm
    deg = scatter_add(edge_weight, col, dim=0, dim_size=num_nodes)
  File "/home/byun/anaconda3/envs/pyg_torch/lib/python3.7/site-packages/torch_scatter/scatter.py", line 35, in scatter_add
    return scatter_sum(src, index, dim, out, dim_size)
  File "/home/byun/anaconda3/envs/pyg_torch/lib/python3.7/site-packages/torch_scatter/scatter.py", line 27, in scatter_sum
    return out.scatter_add_(dim, index, src)
RuntimeError: index 6900 is out of bounds for dimension 0 with size 6900

I think that this problem comes from the torch_scatter.
Could you tell me what I'm missing? How can I solve this problem??

Best regards :)

[rusty1s]

I believe your edge_index does not contain valid indices only, i.e. it contains indices greater than or equal to 6900. In general, you want to ensure that the following assert does pass:

assert data.edge_index.min() >= 0 and data.edge_index.max() < data.num_nodes