A Demonstration of Interpretability Methods for Graph Neural Networks