Code for "Bridging the Gap between Training and Inference: Multi-Candidate Optimization for Diverse Neural Machine Translation" (NAACL 2022) - View it on GitHub