Comment by @punkess • Hey
thanks for sharing, is the model you build trained /specialised on Argentinian literature or will it be able to write English news articles as well? #boo
Stats
Actions: 1
Comments: 0
Likes: 12
Mirrors: 1
Quotes: 0
Comments
Im training it on Latin American literature from scratch, so it should not be able to generalize on other languages. This is because i'm doing it as a means of learning, as well as playing. The goal is to first have it to learn enough so that it knows how to write stories in Latin American style, and then fine tuine it on a specific author. Once I get it to work fully, it should be easy to use other datasets to retrain from scratch with other languages. Im using a Byte-pair encoding tokenizer (OpenAI's TikToken) so it should be relativelly easy to train from scratch with English data.