bugfix: fix the enum/int type mismatch mentioned in #2507

[yzh119]

📌 Description

As mentioned in #2507, trtllm_fp8_per_tensor_scale_moe function would fail when passed integer activation_type.
This PR fixes the type mismatch.

🔍 Related Issues

🚀 Pull Request Checklist

Thank you for contributing to FlashInfer! Before we review your pull request, please make sure the following items are complete.

✅ Pre-commit Checks

• I have installed pre-commit by running pip install pre-commit (or used your preferred method).
• I have installed the hooks with pre-commit install.
• I have run the hooks manually with pre-commit run --all-files and fixed any reported issues.

If you are unsure about how to set up pre-commit, see the pre-commit documentation.

🧪 Tests

• Tests have been added or updated as needed.
• All tests are passing (unittest, etc.).

Reviewer Notes

Summary by CodeRabbit

• Refactor

  • Public APIs updated: activation_type now accepts integer values (defaults adjusted to numeric activation codes).
  • Call sites and public function signatures aligned to use the numeric activation_type.

• Tests

  • Test inputs updated to supply numeric activation_type values instead of enum members.

[gemini-code-assist]

Summary of Changes

Hello @yzh119, 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 resolves a reported bug concerning a type mismatch within the trtllm_fp8_per_tensor_scale_moe function. By ensuring the activation_type parameter is correctly cast to its expected enum type, the change prevents runtime failures and enhances the function's reliability when handling different input types.

Highlights

• Type Mismatch Fix: Corrected an issue in the trtllm_fp8_per_tensor_scale_moe function where passing an integer activation_type would cause a failure. The fix involves explicitly casting the activation_type parameter to ActivationType before passing it to the underlying implementation.

🧠 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

• flashinfer/fused_moe/core.py
  • Modified the call to _trtllm_fp8_per_tensor_scale_moe_impl to explicitly wrap the activation_type argument with ActivationType() to resolve a type mismatch.

Activity

• The author has confirmed completion of pre-commit checks.
• Tests have been added or updated as needed, and all tests are passing.

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. ↩

[coderabbitai]

📝 Walkthrough

Walkthrough

Replaced enum-typed ActivationType parameters with integer activation_type (using .value) across fused MoE public APIs and call sites in flashinfer/fused_moe/core.py, and updated tests to pass enum .value integers accordingly.

Changes

Cohort / File(s)	Summary
Fused MoE core flashinfer/fused_moe/core.py	Changed public function/constructor signatures and call sites to accept activation_type: int (wired to ActivationType.Swiglu.value) instead of ActivationType; replaced uses of activation_type.value with activation_type when forwarding to C++/autotuning paths.
Tests tests/moe/test_trtllm_gen_fused_moe.py	Updated parametrization to pass ActivationType enum .value integers instead of enum members so tests match new int-based API expectations.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• feat: Support Fused MoE non gated Relu2 NVFP4 & FP8 and support Nemotron #2304 — touches activation-type propagation and fused MoE changes similar to this PR.
• Feature: Support non-gated activation in cutlass fused MoE nvfp4 #2011 — adjusts activation_type propagation/representation across fused MoE call sites.
• feat: Support Fused MoE non gated Relu2 NVFP4 & FP8 and support Nemotron, fixed #2462 — related work replacing/renaming gated activation enums and threading activation_type through APIs.

Suggested labels

run-ci

Suggested reviewers

• djmmoss
• cyx-6
• jiahanc
• nv-yunzheq
• IwakuraRein

Poem

🐰 I hopped through code both day and night,
Replaced the enum with a number light,
Values travel clean, no wrapper in sight,
The fences lowered, the logic tight.

🚥 Pre-merge checks | ✅ 2 | ❌ 2

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 57.14% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.
Merge Conflict Detection	⚠️ Warning	❌ Merge conflicts detected (43 files): ⚔️ benchmarks/bench_cute_dsl_blockscaled_gemm.py (content) ⚔️ benchmarks/bench_trtllm_gen_fused_moe_autotuner.py (content) ⚔️ benchmarks/routines/flashinfer_benchmark_utils.py (content) ⚔️ benchmarks/routines/gemm.py (content) ⚔️ csrc/flashinfer_sampling_binding.cu (content) ⚔️ csrc/gdn_prefill_launcher.cu (content) ⚔️ csrc/nv_internal/tensorrt_llm/kernels/communicationKernels/moeAlltoAllKernels.cu (content) ⚔️ csrc/sampling.cu (content) ⚔️ csrc/trtllm_fmha_kernel_launcher.cu (content) ⚔️ docker/Dockerfile.cu126 (content) ⚔️ docker/Dockerfile.cu128 (content) ⚔️ docker/Dockerfile.cu129 (content) ⚔️ docker/Dockerfile.cu130 (content) ⚔️ flashinfer/__init__.py (content) ⚔️ flashinfer/aot.py (content) ⚔️ flashinfer/artifacts.py (content) ⚔️ flashinfer/cute_dsl/__init__.py (content) ⚔️ flashinfer/cute_dsl/blockscaled_gemm.py (content) ⚔️ flashinfer/cute_dsl/utils.py (content) ⚔️ flashinfer/decode.py (content) ⚔️ flashinfer/fused_moe/__init__.py (content) ⚔️ flashinfer/fused_moe/core.py (content) ⚔️ flashinfer/gemm/__init__.py (content) ⚔️ flashinfer/gemm/gemm_base.py (content) ⚔️ flashinfer/jit/__init__.py (content) ⚔️ flashinfer/jit/gemm/__init__.py (content) ⚔️ flashinfer/jit/gemm/core.py (content) ⚔️ flashinfer/mla.py (content) ⚔️ flashinfer/prefill.py (content) ⚔️ flashinfer/sampling.py (content) ⚔️ flashinfer/triton/__init__.py (content) ⚔️ flashinfer/utils.py (content) ⚔️ include/flashinfer/sampling.cuh (content) ⚔️ include/flashinfer/trtllm/fmha/fmhaKernels.cuh (content) ⚔️ include/flashinfer/trtllm/fmha/fmhaRunnerParams.h (content) ⚔️ include/flashinfer/trtllm/fmha/kernelParams.h (content) ⚔️ scripts/authorized_codeowner.txt (content) ⚔️ scripts/task_run_unit_tests.sh (content) ⚔️ scripts/test_utils.sh (content) ⚔️ tests/attention/test_trtllm_gen_attention.py (content) ⚔️ tests/gemm/test_bmm_fp8.py (content) ⚔️ tests/gemm/test_cute_dsl_blockscaled_gemm.py (content) ⚔️ tests/moe/test_trtllm_gen_fused_moe.py (content) These conflicts must be resolved before merging into main.	Resolve conflicts locally and push changes to this branch.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely identifies the main purpose: fixing an enum/int type mismatch in a specific function, with reference to the related issue #2507.
Description check	✅ Passed	The PR description addresses the core issue, provides context from the related issue, and completes most checklist items. However, 'All tests are passing' is explicitly unchecked despite test updates being made.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

⚔️ Resolve merge conflicts (beta)

• Auto-commit resolved conflicts to branch hotfix-2507
• Post resolved changes as copyable diffs in a comment

---

No actionable comments were generated in the recent review. 🎉

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[gemini-code-assist]

Code Review

This pull request addresses a type mismatch issue in the trtllm_fp8_per_tensor_scale_moe function where an integer activation_type was passed to a function expecting an ActivationType enum. The change correctly wraps the integer value with the ActivationType enum constructor, resolving the bug. The fix is concise and accurate. I've reviewed other related functions and they do not appear to have the same issue. The change is approved.

[dbari]

I adapted the tests to match the function signature and made it consistent with the fp4 functions here:
https://github.com/dbari/flashinfer/tree/hotfix-2507

Feel free to include it in this PR, I didn't want to open a new one for the same thing to avoid confusion.

[aleozlx]

lgtm

[IwakuraRein]

trtllm_fp4_block_scale_moe_op has activation_type: int = ActivationType.Swiglu.value,. Maybe it's better to unify trtllm_fp4_block_scale_moe_op and trtllm_fp8_per_tensor_scale_moe

[yzh119]

Hi @dbari I have merged your commits.

Maybe it's better to unify trtllm_fp4_block_scale_moe_op and trtllm_fp8_per_tensor_scale_moe

@IwakuraRein Can you work on a follow up PR? This one might be merged into v0.6.4 as a hotfix.