simple-llm-finetuner
simple-llm-finetuner copied to clipboard
Finetuning in unsupported language
My language was not on the list of 20 languages the original model was trained on. Is it possible to finetune llama with a dataset in a language that was not included in the base model?
I tried finetuning GPT2 which apparently was only trained with english language with german input text and the result is bad. I guess it takes a whole lot of more training in order to generate a new language.
In short: I think you probably need another model
You can absolutely do this. Tell me how it goes!