Build A Reasoning Model From Scratch 7: Reinforcement Learning 2 (Tracking, KL Terms, And More)

This second video on reinforcement learning adds clipped policy rations, a KL loss term, format rewards, and analyzes various training metrics. Plus, it gives additional paper references for more tips and tricks. - Reasoning playlist: https://www.youtube.com/watch?v=Kh9mqTzjuEQ&list=PLfrHAQCRxtUs - Reasoning Book: https://amzn.to/4aAKiFY - Reasoning GitHub repo: https://github.com/rasbt/reasoning-from-scratch - LLMs from Scratch book: https://amzn.to/4fqvn0D - LLMs from Scratch repo: https://github.com/rasbt/LLMs-from-scratch - LLMs from Scratch playlist: https://www.youtube.com/watch?v=yAcWnfsZhzo&list=PLTKMiZHVd_2IIEsoJrWACkIxLRdfMlw11&index=1 00:00 Introduction and recap 01:52 Interpreting basic GRPO training metrics 06:34 Planned improvements to GRPO 08:58 Running longer training jobs with Python scripts 13:39 Running the baseline GRPO training script 17:29 Loading and plotting training logs 19:29 Diagnosing unstable training 23:55 Evaluating checkpoints on MATH-500 26:26 Downloading existing checkpoints 30:09 Tracking advantage statistics 34:53 Understanding entropy 40:32 Computing entropy in PyTorch 44:17 Interpreting entropy values 48:58 Adding entropy tracking to GRPO 53:36 Analyzing advantage and entropy metrics 56:18 Stabilizing GRPO with clipped policy ratios 1:03:27 Implementing the clipped policy loss 1:09:39 Analyzing clipped policy training results 1:11:25 KL divergence and reward hacking 1:15:12 Adding a KL loss term 1:20:34 Limitations of the simplified KL loss 1:23:04 Format rewards and think tags 1:25:47 Adding special tokens to the tokenizer 1:30:29 Implementing the format reward 1:35:56 Analyzing format reward training 1:38:25 Rewarding format only for correct answers 1:40:48 Further GRPO improvements from research 1:45:43 Next steps and distillation