A problem about dropout_adj

[fxsxjtu]

When I try to use the dropout_adj , I encoutered a basic problem.
adj, features, labels, idx_train, idx_val, idx_test = load_benchmark("cora") a,b = dense_to_sparse(adj) a,b = dropout_adj(edge_index=a,p=0.95,edge_attr=b) ab = to_dense_adj(edge_index=a,edge_attr=b)
I just want to dropout some edges in adj , but the size of the adj is also changed!
If there is any way to keep the size while dropping the edges?I'd appreciate some help.

[wsad1]

What might be happening is that dropout_adj is dropping edges between the last few nodes, and when you call to_dense_adj its not able to infer the number of nodes in the graph correctly. Try setting the batch or max_num_nodes parameter in to_dense_adj.

[fxsxjtu]

Thanks alot ,it does work!