I am trying to understand how to do training of multilingual data.
Do we need to create a dataset having text in different languages and their labels? Or there is some sort of settings/flags in bert to let bert know this is multilingual? I mean do we need to provide language code, or bert automatically understand the language name