Hey, I am currently fine tuning a llama 2 model with a model I have set up with JSON function calls. I am wondering if there's any differences in the way I should approach fine tuning with the structured output versus the unstructured output. &#…