Does GCNConv take duplicate edges into account?

[ZRH0308]

Hi,

I am new to geometrics, I need your help with confusion that may be stupid but important to me.

I was wondering when I give GCN.forward() an input like below,

edge_index = torch.tensor([[0, 1, 1, 1, 1, 0],
                           [1, 2, 2, 2, 0, 1]], dtype=torch.long)

x = torch.tensor([[-1], [0], [1]], dtype=torch.float)

How will GCNConv( x, edge_index) process the duplicate edges in edge_index? Merge them or just drop the duplicates?
In other words, in its view, my input will be treated as which of the following matrices? ( I don't add self-loop here)

tensor([[[0., 2., 0.],
         [1., 0., 3.],
         [0., 0., 0.]]])

OR

tensor([[[0., 1., 0.],
         [1., 0., 1.],
         [0., 0., 0.]]])

？

Thank you very much !

[rusty1s]

We will not remove duplicate edges, but instead interpret them as two edges. As such, it will be treated as

tensor([[[0., 2., 0.],
         [1., 0., 3.],
         [0., 0., 0.]]])

[ZRH0308]

Thanks for your reply !!