medibot_chat.csv file contains the the data that can be directly used for training LlaMA2 chat model. The dataset is in the following format:
<s>[INST]{user_query 1}[/INST]{chatbot_response 1}[INST]{user_query 2}[/INST]{chatbot_response 2}....[INST]{user_query n}[/INST]{chatbot response n}</s>
NOTE: The dataset was generated with the help of chatGPT 3.5 and may not be up to medical standards and is definitely not verified. So don't integrate this model for any real-life medical applications. You can of course create a better datasets that is correct and verified and then integrate into some medical applications. This dataset on the other hand is meant only for training the LlaMA2 chat model and see how good the finetuning works.