Memmory error

[m-idasa]

I try to use this community detection implementation for a graph with 6071489 nodes, 52248 indicators, and no embedding dimensions. I try to run it with google collab 12.5 GB ram but I get
Unable to allocate 258. TiB for an array with shape (5949087, 5949087) and data type float64
can you help me how to fix it?

[m-idasa]

Is it possible to use sparse instead? Does torch support all required functions with sparse matrix input?

[HongtengXu]

Is it possible to use sparse instead? Does torch support all required functions with sparse matrix input?

Our current implementation does not support sparse inputs. The GWF model is more suitable for cases with many small graphs rather than those with a single huge graph.