Open marco-c opened 1 year ago
Since we last trained our models, newer and larger datasets have been released. We should re-train them (possibly after fixing a few other quality bugs).
Also, we should try using more monolingual data for back translations. I think we can do this all together.
Could you file a separate issue for that for tracking purposes?
Sure, I added #181
Since we last trained our models, newer and larger datasets have been released. We should re-train them (possibly after fixing a few other quality bugs).