cugraph
cugraph copied to clipboard
rapidsai
cuGraph - RAPIDS Graph Analytics Library
### Is this a new feature, an improvement, or a change to existing functionality? New Feature ### How would you describe the priority of this feature request High ### Please...
Currently, regardless of input type of the edge indices, cugraph-pyg will convert to a numpy array, and then later on, copy the numpy array back to the device. This can...
We need to add a second form of symmetrization for compatibility with GNN use cases, as well as standardize the current behavior when there are edge ids and edge types...
#3789 resolved an issue where empty minibatches were dropped from the bulk sampler. The fix for this problem results in batch ids that may not match up with those provided...
Will help bring small graph performance up to par with native PyG using multithreading. Will also make it easier for users to integrate with native PyG, since the interface will...
cuGraph recently refactored its tests to make MG testing easier, and allow for testing MG workflows with only 1 GPU. The test filenames and annotations were also simplified. cuGraph-PyG and...
Right now the number of seeds per call is set to 200,000 by default, which is low and will result in less than optimal performance. The default should at least...
This option was deprecated in 23.10. It should be removed in 23.12.
Currently, the CSR/CSC output from the bulk sampler is unusable since the number of previous destinations does not match the number of sources. This is most likely because something is...
`torch_geometric.EdgeIndex` natively supports caching CSR/CSC formats and is a better maintained alternative to `torch_sparse`. We should migrate to this new class for the outputs of cuGraph-PyG sampling. https://pytorch-geometric.readthedocs.io/en/latest/generated/torch_geometric.edge_index.EdgeIndex.html#torch_geometric.edge_index.EdgeIndex