Looking for the best model for fine-tuning the translation tasks using rare language pairs. — TLDR; M2M performed really better than mT5. Inspiration Two months ago, I started working on Neural Machine Translation (NMT) for low-resource languages Zindi competition. The challenge is to train an NMT model with the highest ROUGE score using a rare language pair: Yoruba to English. Yoruba is a language spoken in…