
Opis
Reinforcement learning (RL) sits at the heart of aligning modern large language models (LLMs), yet many engineers struggle to learn it because of complex math and expensive infrastructure requirements. This book bridges that gap with clear explanations and hands-on projects that show how to train capable LLMs using free compute and efficient techniques.Start with the mathematical foundations of RL, then progress through supervised fine-tuning (SFT), reward modeling, reinforcement learning from human feedback (RLHF) with proximal policy optimization (PPO), direct preference optimization (DPO), and group relative policy optimization (GRPO). Each concept is reinforced with practical coding exercises and real model outputs, helping you understand the theory and its impact on model behavior.
Learn to fine-tune small open models (0.5B–1.5B parameters) with LoRA, mixed-precision training, and memory optimization. Examples are designed for accessible GPU environments, including free Google Colab sessions when available. You'll also explore process reward models, reasoning with GRPO, test-time compute, self-play, agentic systems, and debugging.
Written by senior Google Applied AI engineers, this book focuses on practical techniques used in real-world LLM development.
By the end of this book, you'll be able to implement, evaluate, and troubleshoot LLM alignment workflows while choosing methods that fit your data and compute budget.
*Email sign-up and proof of purchase required