It would be much more efficient to vectorise them and import them to a vector DB. You won’t gain much by fine tuning a LLM on them
good idea, thanks
reply