Would love a breakdown on how this fine-tuning is accomplished. What was the format of the input data? How many prompt/response pairs was it fine-tuned on? Wher

Would love a breakdown on how this fine-tuning is accomplished. What was the format of the input data? How many prompt/response pairs was it fine-tuned on? Where/how did you extract the prompt/response pairs? Great work btw!