Support for new languages

[pyai88]

Hello, thank you for the great work. I have a couple of questions:

1. Should this method work on a new language out-of-the-box? I'm seeing properties from the training language in my output, so I'm wondering if I've made a mistake.
2. If it doesn't work out-of-the-box, would fine-tuning the pre-trained model be preferable to training from scratch?
3. To help me estimate the training cost, could you provide guidance on how much data is typically needed for a new language and how long you trained your pre-trained model for?

Thank you in advance.

[Plachtaa]

Hi there,
If you find the output in unseen language being accented, you may try finetuning the current checkpoint with the language you desire.
I cannot give an estimation how many hours of data is required, the only thing I suggest is to use as much as you have