1.4.3

Features

• Cluster-GCN via ClusterData and ClusterLoader for operating on large-scale graphs, see examples/cluster_gcn.py for an example on how to use
• Added a tutorial about advanced mini-batching scenarios
• Added a tensorboard logging example

Datasets

• CitationFull: The full citation network dataset suite
• SNAPDataset: A subset of graph datasets from the SNAP dataset collection
• SuiteSparseMatrixCollection
• TrackMLParticleTrackingDataset

Minor Changes

• Added the concat argument to SAGEConv
• Outsourced the train_test_split_edges method of the graph autoencoder GAE class to torch_geometric.utils

Bugfixes

• Fixed SplineConv compatibility with latest torch-spline-conv package
• trimesh conversion utilities do not longer result in a permutation of the input data