How could I compute the powers of adjacency matrix efficiently in PyG?

[Kousaka-Honoka]

I didn't see the multiplication of sparse tensor in pytorch document, so are there any approaches in PyG that could I implement this operation?
AND I used torch.sparse.mm() to compute A@A, but it gives me a ridiculous bug which is 'SparseTensor' object has no attribute 'is_sparse'

[Kousaka-Honoka]

Well, the SparseTensor in torch_sparse is different from torch.sparse.tensor, and I found It supports @ operator, which can compute the powers of adjacency matrix iteratively

[rusty1s]

Yes, the SparseTensor of torch-sparse supports sparse-sparse matrix multiplication via @. Keep in mind that this will increase the number of non-zero elements in your graph by a large margin :)