torch_geometric.loader clusterData Issue with Partitioning
See original GitHub issueI’m looping through and trying to cluster large graphs into smaller partitions. When I use the clusterData function like so:
graph = pytorch_geometric.Data.data object
graph2 = pytorch_geometric.Data.data object
clusters1 = ClusterData(graph, 200)
clusters2 = ClusterData(graph2,200)
etc. (I’m doing much more than 2 graphs but it’s the same code repeated in a loop)
The code works for some of the graphs and I get the expected output:
Computing METIS partitioning...
Done!
but then half way through I get this error:
Process finished with exit code -1073741819 (0xC0000005)
I’ve gotten a similar partitioning error across both a Mac and a windows 10 laptop. If someone could help me out with this or at least point me to some resources where I might go about solving this issue it would be so helpful.
Environment: python=3.7.11 torch==1.9.0 torch_cluster== 1.5.9 torch_geometric== 2.0.1 torch_scatter== 2.0.8 torch_sparse== 0.6.12 torch_spline_conv== 1.2.1 torchmetrics== 0.5.1
Issue Analytics
- State:
- Created 2 years ago
- Comments:18 (8 by maintainers)
Top GitHub Comments
I see. You can copy the
ClusterData
class and apply some modifications to it (adding self-loops, adding reverse edges), e.g.:which jupyter kernerl version are you using?