Mastering Generative AI: Advanced Fine-Tuning for LLMs

This intermediate-level course addresses the high demand for AI engineers skilled in advanced fine-tuning techniques for large language models (LLMs). Designed for AI specialists looking to enhance their career prospects, the curriculum focuses on cutting-edge methods to customize pre-trained models for specific applications like chatbots, translation, and content generation. Students will explore instruction-tuning with Hugging Face, reward modeling, and training specialized reward models to evaluate LLM outputs. The course delves into reinforcement learning from human feedback (RLHF), teaching participants how to view LLMs as probability distributions and policies that can be optimized. Key techniques covered include proximal policy optimization (PPO) with Hugging Face configuration and direct preference optimization (DPO) using partition functions. Through hands-on labs, students gain practical experience implementing reward modeling, PPO, and DPO approaches. This focused two-week program delivers the specialized knowledge employers are actively seeking, enabling AI professionals to tailor large language models with sophisticated techniques that improve accuracy and relevance in real-world applications.