I want to generate an instruct dataset in JSON format from a text file. Most available resources online already assume that I have the JSON format already and want to fine-tune an LLM. However, I only have raw text files. Is there a way to efficiently generate an instruction dataset from a raw text file in Python?