From molecules to PyTorch Tensors
In this lesson we look at how to go from graphs representing a problem all the way to PyTorch tensors. We’ll focus on creating a Dataset and how to pack multitple graph tensors in to a minibatch. We will use molecular graphs as a practical examples, since they are intuitively represented by graphs and allow us to work with real data for graph prediction tasks.
The lesson ends with trying out a Graph Neural Network on a molecular prediction task without going in to the details of the network architecture.