A bug found in SGA on Pubmed dataset

[benjaminnNgo]

Hi!
Thanks for a very nice package.

I am having an error while running the example, provided inside the SGA class, of the SGA attack on the Pubmed dataset. Here are the code chunk and the error I got:

self.selfloop_degree = torch.tensor(adj.sum(1).A1 + 1, device=self.device) .AttributeError: 'numpy.ndarray' object has no attribute 'A1'