shape matching issue in set_diag

[allenhaozhu]

It seems like pytorch_sparse issue:
Traceback (most recent call last):
File "preprocessing.py", line 142, in
main()
File "preprocessing.py", line 100, in main
adj = get_adj(row, col, N, asymm_norm=args.undirected_asymm_norm, set_diag=args.undirected_set_diag, remove_diag=args.undirected_remove_diag)
File "preprocessing.py", line 19, in get_adj
adj = adj.set_diag()
File "/home/zhu044/.local/lib/python3.6/site-packages/torch_sparse/diag.py", line 96, in
self, values, k)
File "/home/zhu044/.local/lib/python3.6/site-packages/torch_sparse/diag.py", line 47, in set_diag
new_row[mask] = row
RuntimeError: shape mismatch: value tensor of shape [50238473] cannot be broadcast to indexing result of shape [50056641]

[rusty1s]

Can you give a little more information of the adj that produces this error? Do you have a simple example to reproduce?

[allenhaozhu]

Can you give a little more information of the adj that produces this error? Do you have a simple example to reproduce?

parser = argparse.ArgumentParser(description='OGBN-papers100M (MLP)')
parser.add_argument('--num_propagations', type=int, default=3)
parser.add_argument('--dropedge_rate', type=float, default=0.3)
args = parser.parse_args()

# SGC pre-processing ######################################################

dataset = PygNodePropPredDataset('ogbn-papers100M')
split_idx = dataset.get_idx_split()
data = dataset[0]

x = data.x.numpy()
N = data.num_nodes

print('Making the graph undirected.')
### Randomly drop some edges to save computation
data.edge_index, _ = dropout_adj(data.edge_index, p = args.dropedge_rate, num_nodes= data.num_nodes)
data.edge_index = to_undirected(data.edge_index, data.num_nodes)

print(data)

row, col = data.edge_index

print('Computing adj...')

adj = SparseTensor(row=row, col=col, sparse_sizes=(N, N))
adj = adj.set_diag()

I also tried SIGN code on papers100, the same problem happend

[Gmrylbx]

Has this bug been fixed? I also encountered a similar problem

[rusty1s]

It works fine for me on my end. What's the problem you encounter?

[hieuvt29]

I encountered the same issue and after a quick check, I found that it happens because this function (https://github.com/rusty1s/pytorch_sparse/blob/master/torch_sparse/diag.py#L32) assumes that the columns are sorted but it is not true for my case. Best!