The code for our paper in EMNLP2023: Continual Learning for Multilingual Neural Machine Translation via Dual Importance-based Model Division