Anyone know how to train Marian MT from scratch? I am trying to train Marian MT for English to Korean, But I stuck on getting the right tokenizer, I tried using the Korean to English tokenizer but it resulted with error, The code I was trying is an example from here Google Colaboratory.
Hello.
Unfortunately, this is not a response to your request. I’m trying to train the same model for a language pair translation that doesn’t exist in hugginface. So the only way to go is to train it from scratch.
Did you figure out yet how to do it?