SMILe: NeurIPS Challenge 2019 Reproducing SMILe: Scalable Meta Inverse Reinforcement Learning through Context-Conditional Policies (Zemel et al.).