How to Align Large Language Models with Human Preferences Using Direct Preference Optimization, QLoRA, and Ultra-Feedback