pytorch_geometric/examples /autoencoder.py hardcodes the outchannel to 16 the line reads:
in_channels, out_channels = dataset.num_features, 16
possible datasets for this file are 'Cora', 'CiteSeer', 'PubMed' is the assignment of out_channels the number of predictions based on the number of labels? If the dataset had binary labels, should out_channels be 2?