Reinforcement Learning with GRPO, Clearly Explained!!!
From StatQuest with Josh Starmer · · 26 min

The player loads from YouTube only when you click. It uses YouTube's privacy-enhanced mode.
If the video does not play, the owner may have turned off embedding. Watch it on YouTube instead.
Description
In 2024, the GRPO method for training Reasoning Models took the world by storm by simplifying the Reinforcement Learning stage of aligning a Large Language Model. In this StatQuest, we go through the GRPO loss function one step at at time, learning exactly what it does and how it does it. BAM!
NOTE: This StatQuest assumes that you are already familiar with... The Main Ideas of Neural Networks: https://youtu.be/CqOfi41LfDw Backpropagation: https://youtu.be/IN2XmBhILt4 Decoder-Only Transformers: https://youtu.be/bQ5BoolX9Ag
If you'd like to support StatQuest, please consider... Patreon: https://www.patreon.com/statquest ...or... YouTube Membership: https://www.youtube.com/channel/UCtYLUTtgS3k1Fg4y5tAhLbw/join
...buying a book, a study guide, a t-shirt or hoodie, or a song from the StatQuest store... https://statquest.org/statquest-store/
...or just donating to StatQuest! paypal: https://www.paypal.me/statquest venmo: @JoshStarmer
Lastly, if you want to keep up with me as I research and create new StatQuests, follow me on twitter: https://twitter.com/joshuastarmer
0:00 Awesome song and introduction 1:00 The main ideas of GRPO 6:18 A few details about GRPO 8:15 GRPO, one step at a time 23:28 Summary 24:20 The KL Divergence term
#StatQuest