[Feature] Enable activation selection for fp4 cutlass moe

[yilian49]

Motivation

The current cutlass_moe_fp4 in cutlass_moe.py hardcodes silu as the activation and raises an error for any other value. Since the file already includes implementations for gelu_tanh and gelu, a simple change can enable support for selecting among these activations.

Modifications

Added an activation string argument to cutlass_moe_fp4, with a check to ensure it is one of gelu_tanh, silu, or gelu. The corresponding kernel is invoked based on the selection.

Extension

The flashinfer moe directly calls the flashinfer library, so modification to include activation selection is left for future work.

Checklist

• Format your code according to the Code Formatting with Pre-Commit.
• Add unit tests as outlined in the Running Unit Tests.
• Update documentation / docstrings / example tutorials as needed, according to Writing Documentation.
• Provide throughput / latency benchmark results and accuracy evaluation results as needed, according to Benchmark and Profiling and Accuracy Results.
• For reviewers: If you haven't made any contributions to this PR and are only assisting with merging the main branch, please remove yourself as a co-author when merging the PR.
• Please feel free to join our Slack channel at https://slack.sglang.ai to discuss your PR.

[gemini-code-assist]

Summary of Changes

Hello @yilian49, 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 enhances the flexibility of the FP4 Mixture-of-Experts (MoE) layer by enabling the selection of different activation functions. Previously, the cutlass_moe_fp4 implementation was restricted to the SiLU activation. This change allows users to choose between SiLU, GeLU, and GeLU-Tanh, providing more architectural options for models utilizing FP4 MoE.

Highlights

• Configurable Activation Functions: The cutlass_moe_fp4 function now supports selecting between "silu", "gelu", and "gelu_tanh" activation functions, removing the previous hardcoded "silu" limitation.
• API Changes: The cutlass_moe_fp4 function in cutlass_moe.py now accepts an activation string argument, and this argument is propagated from modelopt_quant.py.
• Test Coverage: Unit tests for cutlass_fp4_moe_no_graph have been updated to include parametrization for "silu", "gelu", and "gelu_tanh" activations, ensuring proper functionality across supported options.

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 in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.

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 issue 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 is currently in preview and 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 to provide feedback.

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 successfully extends the cutlass_moe_fp4 function to support multiple activation functions (gelu, gelu_tanh, and silu) instead of hardcoding silu. The changes are well-implemented, including necessary imports, parameter additions, updated docstrings, and conditional logic to select the appropriate activation kernel. The flashinfer path correctly maintains its silu-only restriction, and the unit tests have been updated to cover the new activation options, ensuring correctness and comprehensive testing.

[gemini-code-assist]

The error message is a bit redundant. Consider making it more concise.

Suggested change

	raise ValueError(
	f"Unsupported activation function: {activation}. "
	"Supported values are 'silu', 'gelu', 'gelu_tanh'."
	raise ValueError(
	f"Unsupported activation function: '{activation}'. "
	"Expected one of 'silu', 'gelu', 'gelu_tanh'."
	)