Mistral/Zephyr 7B fine-tuning! Make LLM-brained robot. Part 1: creating data.

Task of running local LLM as Mistral 7B or Zephyr 7B on Raspberry Pi to make it a brain of autonomous self-thinking robot is very challenging. Single board computers as Raspberry Pi have very limited resourses, and LLMs, otherwise, usually very resourse-consuming. Another problem here is that small 7B LLMs are not smart enough to realize complex tasks. To solve that problems we'll use procedure called fine-tuning: we'll teach our LLM to be good at being brain of our autonomus robot. That will increase intelligence of the model in that particular task and could increase it's efficiency as less …