docs: add Important Metrics to Monitor section#894
Conversation
Add comprehensive metrics monitoring guide with three subsections: - Reward Metrics: eval-rollout/reward, rollout/reward, ppo/actor/task_reward with troubleshooting for high variance - Importance Weight Metrics: Explain decoupled PPO loss formulation with importance_weight and behave_imp_weight, including troubleshooting tips - Sequence Length Metrics: no_eos_ratio and seq_len for detecting truncation Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Summary of ChangesHello @xssstory, 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 reinforcement learning performance diagnostics documentation by integrating a comprehensive 'Important Metrics to Monitor' section. This new content aims to equip users with essential knowledge and troubleshooting steps for key metrics across reward, importance weight, and sequence length categories, ultimately fostering more stable training and earlier issue detection. 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. Changelog
Activity
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
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request adds a valuable 'Important Metrics to Monitor' section to the RL performance diagnostics guide. The new content is well-structured, providing clear explanations of key metrics for rewards, importance weights, and sequence length, along with practical troubleshooting advice. My review includes a correction for a typo in a parameter name and a suggestion to clarify a minor naming inconsistency to further improve the documentation's clarity for users.
| - If deviation persists at `ppo_n_minibatches == 1` (common in MoE training), add | ||
| `actor.megatron.use_deterministic_algorithms=1` to your config. | ||
|
|
||
| **Troubleshooting `behave_imp_weight` deviations:** |
There was a problem hiding this comment.
There's a slight naming inconsistency that could be confusing for users. This section and the metric itself are named behave_imp_weight, but the corresponding configuration parameter is behav_imp_weight_cap. While this reflects the codebase, adding a small note to clarify this difference would enhance the documentation's clarity.
* docs: polish Important Metrics to Monitor section Add comprehensive metrics monitoring guide with three subsections: - Reward Metrics: eval-rollout/reward, rollout/reward, ppo/actor/task_reward with troubleshooting for high variance - Importance Weight Metrics: Explain decoupled PPO loss formulation with importance_weight and behave_imp_weight, including troubleshooting tips - Sequence Length Metrics: no_eos_ratio and seq_len for detecting truncation Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix doc format --------- Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com> Co-authored-by: 博惟 <bowei.fw@antgroup.com>
* docs: polish Important Metrics to Monitor section Add comprehensive metrics monitoring guide with three subsections: - Reward Metrics: eval-rollout/reward, rollout/reward, ppo/actor/task_reward with troubleshooting for high variance - Importance Weight Metrics: Explain decoupled PPO loss formulation with importance_weight and behave_imp_weight, including troubleshooting tips - Sequence Length Metrics: no_eos_ratio and seq_len for detecting truncation Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix doc format --------- Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com> Co-authored-by: 博惟 <bowei.fw@antgroup.com>
2 participants
Description
Add a comprehensive "Important Metrics to Monitor" section to the RL performance
diagnostics guide. This includes three subsections covering reward metrics,
importance weight metrics (with decoupled PPO loss formulation), and sequence
length metrics, along with troubleshooting guidance for each.
Related Issue
N/A
Type of Change
Checklist
jb build docs/gemini review)Breaking Change Details (if applicable):
N/A
Additional Context
Key additions:
eval-rollout/reward,rollout/reward,ppo/actor/task_rewardimportance_weight,behave_imp_weightwith decoupled PPO loss formulano_eos_ratio,seq_lenfor truncation detectionFiles changed:
docs/best_practices/algo_perf.md: Add Important Metrics to Monitor section