RAFT problem

[chahn85]

I am trying to apply RAFT to a set of medical guidelines. I applied the raft dataset generation notebook (https://github.com/run-llama/llama_index/blob/main/llama-index-packs/llama-index-packs-raft-dataset/examples/raft_dataset.ipynb) to generate the dataset but realized that there was nowhere to enter the "distractor documents". How should I set up the pipeline for including the "distractor documents"? Is it critical?

Also, when I try to finetune LLAMA2 in AzureAI using the dataset generated from the RAFT dataset generation notebook, I see that I only get to choose 2 columns: "prompt" and "completion". I chose "question" for "prompt" and "cot_answer" for "completion". Is this sufficient? How will the other aspects such as oracle_context get incorporated into finetuning? Thank you for your help.