Synthetic Preference Datasets for Continual Reinforcement Learning from Human Feedback - https://github.com/ComplexData-MILA/AIF-Gen