rather professional readme document

[zhaochenyang20]

No description provided.

[gemini-code-assist]

Summary of Changes

Hello @zhaochenyang20, 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 updates the project's README.md to offer a more structured and informative overview of the Miles reinforcement learning framework. The changes aim to better articulate the framework's enterprise-grade capabilities, advanced features for large-scale MoE models, and its focus on performance and stability. The revised documentation provides clearer guidance for new users and a more transparent view of the project's ongoing development and future plans.

Highlights

• Enhanced Project Overview: The README.md has been completely restructured to provide a clearer and more comprehensive introduction to the Miles framework, including its purpose, key features, and capabilities.
• Detailed Key Features Section: New sections are introduced to highlight advanced functionalities such as Unified FP8 Pipeline, Rollout Routing Replay (R3), INT4 QAT Support, and methods for eliminating train-inference mismatch and improving performance.
• Updated Model Support and Training Scenarios: The documentation now explicitly lists supported model families (DeepSeek, Qwen, Llama, Gemma, etc.) and diverse training scenarios (multi-turn, VLM/LLM, reasoning, multi-agent).
• Refreshed Latest Updates and Roadmap: The 'News' section is renamed to 'Latest Updates' with new, forward-dated entries, and the 'Roadmap' is reorganized into 'Completed' and 'In Progress & Planned' items, reflecting recent achievements and future directions.
• Improved Quick Start Guide: The quick start section now includes clear installation instructions (Docker and source) and a concrete example for launching training.

Using Gemini Code Assist

The 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 /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

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 .gemini/ folder in the base of the repository. Detailed instructions can be found here.

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

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request introduces a significant and professional overhaul of the README.md file. The new version is much better structured, more visually appealing, and provides a clearer and more comprehensive overview of the Miles project. The added sections on model support and the reorganized key features are particularly helpful. I've identified a few minor issues, including a typo and some broken links, which I've detailed in the comments. Overall, this is an excellent update that greatly improves the project's documentation.

[BHZ-BER]

I think Ying also mentioned a few times that truly on-policy is a bit confusing. Shall we change the name a bit? Say "Bit-wise identical training and inference alignment"?

[BHZ-BER]

Can we be a bit more explicit and comprehensive? Say at least we know GLM 4.5, 4.6, 4.7 are supported. What about Minimax M2 / M2.1?

[BHZ-BER]

"any model supported by SGLang and Megatron" maybe we shall also add FSDP here? But also we need to mention that FSDP is still experimental. It may be unstable etc.

[BHZ-BER]

For here is it possible to use a megatron based backend rather than fsdp? Given that fsdp is still a bit experimental and may have some memory issues.

Also at some point we shall say we support both FSDP2 and Megatron training backend, but the FSDP backend is still under development and experimental?