Convert the edge weigth into the edge_attr

[nicolasH1027]

Suppose I have a weighted graph constructed like below

import torch
from torch_geometric.data import Data
from torch_geometric.utils import to_dense_adj

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

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

data = Data(x=x, edge_index=edge_index)

to_dense_adj(edge_index )

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

How can I converted it into a unweighted graph with the original weight as its edge_attr.

In this case, the data would becomes like

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

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

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


data = Data(x=x, edge_index=edge_index, edge_attr = edge_attr  )
to_dense_adj(data['edge_index '])

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

data['edge_index ']

tensor([[2],
        [2],
        [1],
        [1]])

Any help would be appreciated

[wsad1]

Check out dense_to_sparse . It converts a dense adjacency matrix to a sparse adjacency matrix as described here.

[nicolasH1027]

Thank you so much, I didnt notich this function. I make an another way which might be more memory efficient for the whole operation.

[nicolasH1027]

As the @wsad1 pointed out, we can achieve this conviently by using dense_to_sparse, or if the adjacency matrix is very big, then the following method might work.

edge = [tuple(i) for i in edge_index.T.tolist()]
d = dict()
for i in edge:
  if i in d:
    d[i] += 1
  else:
    d[i] = 1
edge = torch.tensor([i for i in d.keys()]).t().contiguous()
weight = torch.tensor(list(d.values())).reshape(-1,1)