Welcome to RL-Insight’s documentation!

RL-Insight is a performance insight and observability toolkit for reinforcement learning (RL) systems. It focuses on analyzing offline cluster logs first, and is evolving towards full-stack online monitoring and visualization.

---

Overview

• Architecture
  • 1. 框架概览
  • 2. 模块简介
• RL Timeline quickstart
  • 1. 简介
  • 2. 安装
  • 3. 快速使用
  • 4. 命令行参数
  • 5. 输出说明
  • 6. 注意事项
• GMM Expert Load visualization quickstart
  • 一、简介
  • 二、快速使用
  • 三、命令行参数
  • 四、输出说明
  • 五、注意事项
• Memory Parser quickstart
  • 1. 简介
  • 2. 输入数据
  • 3. 快速使用
  • 4. 输出说明
  • 5. 局限性

Data Specification

• Data Specification and Format Guide
  • 1. Torch Profiler 数据
  • 2. MSTX（Ascend）Profiling 数据
  • 3. NVTX Profiling 数据
  • 4. 输出生成summary_event数据
  • 5. VeRL 训练日志（可选校验）
  • 5. GMM 专家负载dump数据
  • 6. Ascend Memory Profiling 数据

Utils

• MSTX Preprocessing
  • 1. 功能说明
  • 2. 脚本位置
  • 3. 使用方法
  • 4. 参数说明
  • 5. profile目录结构示意
  • 6. 注意事项

Developer Guides

• Extending Guide
  • 1. 扩展 DataRule
  • 2. 扩展 Parser / Visualizer
  • 3. 扩展 Pipeline
• BaseClusterParser Interface
  • 1. run() 执行入口
  • 2. allocate_prof_data()
  • 3. parse_analysis_data()
• DataRule Extending Guide
  • 1. 实现自定义 DataRule
  • 2. 注册自定义DataRule
• Memory Parser Guide
  • 1. 设计概述
  • 2. 内部接口
  • 3. 扩展指南
  • 4. 依赖
  • 5. 测试

Contribution

RL-Insight is free software; you can redistribute it and/or modify it under the terms of the Apache License 2.0. We welcome contributions. Join us on GitHub for discussions.

Contributions from the community are welcome! Please check out our project roadmap and good first issues to see where you can contribute.

Code Linting and Formatting

We use pre-commit to help improve code quality. To initialize pre-commit, run:

pip install pre-commit
pre-commit install

To resolve CI errors locally, you can also manually run pre-commit by:

pre-commit run

Adding CI tests

If possible, please add CI test(s) for your new feature:

1. Find the most relevant workflow yml file.

2. Add related path patterns to the paths section if not already included.

3. Minimize the workload of the test script(s) (see existing scripts for examples).