Revert "[Blackwell] Make mxint4 flashinfer_trtllm moe gemm set by default on blackwell"

[vincentzed]

Reverts #18136

[gemini-code-assist]

Summary of Changes

Hello, 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 undoes a recent modification that introduced an automatic configuration for the MoE runner backend based on specific int4 quantization parameters for certain models on Blackwell architecture. The revert streamlines the backend selection logic, removing the special handling for Kimi K2/K2.5 int4 models and maintaining the existing default behavior for DeepseekV3ForCausalLM with FP8/modelopt quantization.

Highlights

• Reverted MoE Backend Default: The pull request reverts a previous change that automatically set the flashinfer_trtllm MoE runner backend for Kimi K2/K2.5 models utilizing a specific int4 quantization on Blackwell (sm100) GPUs.
• Removed Int4 Detection Logic: The logic used to detect is_kimi_k2_k25_thinking_int4 based on quantization configuration has been removed, simplifying the model-specific adjustment process.

🧠 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

• python/sglang/srt/server_args.py
  • Removed the is_kimi_k2_k25_thinking_int4 variable definition and associated logic.
  • Simplified the conditional statement for setting self.moe_runner_backend by removing the is_kimi_k2_k25_thinking_int4 check.
  • Updated the logging message to reflect the removal of Kimi K2/K2.5 specific int4 handling.

Activity

• No activity has been recorded for this pull request yet.

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 reverts a previous change that added default support for mxint4 flashinfer_trtllm moe gemm on Blackwell hardware. The revert cleanly removes the logic for detecting a specific int4 quantization configuration and the associated conditional logic for setting the MoE runner backend. The code is restored to its state prior to the reverted change.

[JustinTong0323]

fixed in #19696