How do I convert a base model to an instruct model? so far I tried unsloth and ms-swift and it didn’t work.
llama3.1 8b base → instruct
I know some instruct fine tuning is involved but then the model that is produced won’t have <|eot_id|> as eos or correct chat template in the tokenizer_config.json… I manually edited the files but it doesn’t seem to work.
I achieved proper answers but it doesn’t seem to end with <|eot_id|>. Do I have to fine tune for days and days to see it properly end?