Improving LLMs with RLHF
Goals: Equip students with knowledge and practical skills for implementing RLHF.
This short module is focussed on Reinforcement Learning from Human Feedback (RLHF). We learn how to incorporate human feedback into the training process through a reward model that learns the desired patterns to amplify the model’s output.
- Deep Dive into RLHF: This lesson explores the mechanics and applications of RLHF. We will provide a robust understanding of how RLHF functions and its significance in LLM training and optimization.
- Improving trained models with RLHF: This lesson provides a practical guide on RLHF as a fine-tuning technique for LLMs. We build upon our previous fine tuning example by implementing RLHF.