Translation script for datasets using normistral-7b-warm?
Hi!
Thank you for all the great work you are doing! This is fantastic! Keep it up!
Would you mind sharing the code (or general snippets) for the translation of the datasets you mention in the model card? I've been using Meta's SeamlessM4T for translating datasets from english to norwegian, but it's kind of slow and the translation isn't allways complete. I would really like to test normistral-7b-warm for the job.
Hi, we used the same few-shot setup as in the English->Bokmål/Nynorsk evaluation. This prompt template: {source_language}: {source_text}\n{target_language}: where source_language="Engelsk" and target_language="Bokmål", for example. Then we used 6-shot setup with manually written in-domain translations; each shot is separated by a double-newline in the prompt.