chore: minor fix doc formula#793
Conversation
Summary of ChangesHello @ZiyiTsang, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request significantly enhances the project's documentation by introducing a dedicated page for the foundational PPO algorithm, correcting a critical mathematical formula in the GRPO documentation, and standardizing parameter headings across various algorithm descriptions. These changes aim to improve the accuracy, clarity, and navigability of the technical documentation for reinforcement learning algorithms. Highlights
🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console. Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
ZiyiTsang
left a comment
ZiyiTsang
left a comment
There was a problem hiding this comment.
/gemini review
![gemini-code-assist[bot]](https://avatars.gh.lumenfield.work/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request provides several documentation improvements, primarily fixing the GRPO formula and adding a new page for PPO. The changes to the GRPO formula are a good step, and I've suggested one more refinement to improve its clarity and accuracy with respect to the GRPO algorithm. The other changes, such as adding a PPO documentation page and making headings consistent, are also beneficial.
|
|
||
| $$ | ||
| J_{\text{GRPO}}(\theta) = \mathbb{E}_{\substack{q \sim P(Q),\\ \{o_i\}_{i=1}^G \sim \pi_{\theta_{\text{old}}}(O|q)}} \left[ \frac{1}{G} \sum_{i=1}^G \frac{1}{|o_i|} \sum_{t=1}^{|o_i|} \min\left( r_{i,t}(\theta) \hat{A}_{i,t},\ \text{clip}\left( r_{i,t}(\theta),\ 1-\epsilon,\ 1+\epsilon \right) \hat{A}_{i,t} \right) - \beta D_{\mathrm{KL}}\left[ \pi_\theta \middle| \pi_{\text{ref}} \right] \right] | ||
| J_{\text{GRPO}}(\theta) = \mathbb{E}_{\substack{q \sim P(Q),\\ \{o_i\}_{i=1}^G \sim \pi_{\theta_{\text{old}}}(O|q)}} \left[ \frac{1}{G} \sum_{i=1}^G \sum_{t=1}^{|o_i|} \min\left( r_{i,t}(\theta) \hat{A}_{i,t},\ \text{clip}\left( r_{i,t}(\theta),\ 1-\epsilon,\ 1+\epsilon \right) \hat{A}_{i,t} \right) - \beta D_{\mathrm{KL}}\left[ \pi_\theta \middle| \pi_{\text{ref}} \right] \right] |
There was a problem hiding this comment.
While the KL divergence term can be part of a PPO-style objective, it's not a core component of GRPO itself. As noted in the parameters section, kl_ctl defaults to 0.0 for GRPO. To make the formula more specific to GRPO and less confusing, I suggest removing the KL divergence term.
| J_{\text{GRPO}}(\theta) = \mathbb{E}_{\substack{q \sim P(Q),\\ \{o_i\}_{i=1}^G \sim \pi_{\theta_{\text{old}}}(O|q)}} \left[ \frac{1}{G} \sum_{i=1}^G \sum_{t=1}^{|o_i|} \min\left( r_{i,t}(\theta) \hat{A}_{i,t},\ \text{clip}\left( r_{i,t}(\theta),\ 1-\epsilon,\ 1+\epsilon \right) \hat{A}_{i,t} \right) - \beta D_{\mathrm{KL}}\left[ \pi_\theta \middle| \pi_{\text{ref}} \right] \right] | |
| J_{\text{GRPO}}(\theta) = \mathbb{E}_{\substack{q \sim P(Q),\\ \{o_i\}_{i=1}^G \sim \pi_{\theta_{\text{old}}}(O|q)}} \left[ \frac{1}{G} \sum_{i=1}^G \sum_{t=1}^{|o_i|} \min\left( r_{i,t}(\theta) \hat{A}_{i,t},\ \text{clip}\left( r_{i,t}(\theta),\ 1-\epsilon,\ 1+\epsilon \right) \hat{A}_{i,t} \right) \right] |
2 participants
Description
Fix formula in doc about GRPO and add PPO.md
Please help me to check again the GRPO formula.
Related Issue
Fix issue #784.
Fixes #(issue)
Type of Change
work as expected)
Checklist
jb build docs/gemini review)Breaking Change Details (if applicable):
Additional Context
Need help? Check the Contributing Guide or ask in
GitHub Discussions!