Interpretability for sequence generation models 🐛 🔍
Code for our ICML '19 paper: Neural Network Attributions: A Causal Persp...