5 useful things you'll learn in my new post-training textbook (shipping now!)
After a few long years of finding time to document my lessons from training open models, my post-training book is done!
1. Understanding RL algorithms: The book delves into various reinforcement learning algorithms, such as policy-gradient theorem, PPO, GSPO, and CISPO, to help readers comprehend if a new algorithm is genuine or promising. It presents these algorithms in an intuitive manner, aiding readers in grasping their designs and numerical issues.
2. Post-training techniques and trade-offs: The book explains the importance of post-training techniques for aligning and fine-tuning large language models (LLMs). It highlights trade-offs to be considered while executing post-training methods and dispels common misconceptions.
3. High voice explanatory text: The content of the book features higher voice explanatory text, aimed at providing a comprehensive understanding of post-training methods for LLMs. This style of text was chosen to mirror the author's perspective, making it easier for readers to follow complex concepts.
4. Online availability and supplemental resources: The book is available online for free along with a 12-hour course that includes slides, a video on YouTube, a codebase, and suggested exercises. The complete package is designed to help readers grasp post-training methods and apply them effectively in their research or projects.
5. Discount offer: The book is currently available at a 50% discount from August 19th onwards on Manning. It is also available on Amazon US and UK, with the US edition shipping from August and the UK edition in October.
Written by urgent.news from Interconnects's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
