GNN notebook errors on eval

[jalving]

The GNN notebook recently starting throwing an error with print(nn1(torch.tensor(X).float(), torch.tensor(edges)).detach().numpy())

Is this something you can take a look at @zshiqiang?

[zshiqiang]

I think it is because that a new version of torch_geommetric refactored Sequential. I was using (pooling, "x, None -> x") to avoid the use of batch, which is not working now. So I add batch back as (pooling, "x, batch -> x") and it works for both older version and the lastest version. The notebook is prepared and I will pull a request for it.

[jalving]

Fixed by #167. Thanks @zshiqiang!