First attempt at supervised finetuning using kastan/rlhf-qa-conditional-generation-v2

initial commit