[Relax] Fix HardSigmoid returns 1.0 for NaN input

[cchung100m]

Hi Commiters,

This PR is trying to fix issues #17999. Any suggestions would be appreciated if you are available.

Solutions:

• Compute the clipped result as before
• Check for NaN values using relax.op.isnan(x)
• Use relax.op.where to return NaN for positions where input is NaN, otherwise return the clipped value

[gemini-code-assist]

Summary of Changes

Hello @cchung100m, 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 addresses a critical bug in the Relax framework's HardSigmoid operator where NaN inputs were not handled correctly, leading to an erroneous output of 1.0. The changes ensure that HardSigmoid now propagates NaN values as expected, improving the numerical stability and correctness of computations involving undefined or unrepresentable floating-point numbers. The fix is accompanied by a dedicated test case to validate the new behavior.

Highlights

• HardSigmoid NaN Handling: The HardSigmoid operator in relax.frontend.onnx has been updated to correctly propagate NaN (Not a Number) values. Previously, NaN inputs would incorrectly result in an output of 1.0.
• Implementation Details: The fix involves computing the standard clipped result, then explicitly checking for NaN inputs using relax.op.isnan. If an input is NaN, relax.op.where is used to return a NaN value (generated by 0.0 / 0.0) at that position, otherwise the clipped value is returned.
• New Test Case: A new test, test_hardsigmoid_nan, has been added to tests/python/relax/test_frontend_onnx.py to verify that HardSigmoid correctly preserves NaN values in its output when given NaN inputs.

🧠 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.

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 addresses an issue where HardSigmoid would return 1.0 for a NaN input, instead of propagating the NaN. The fix correctly uses relax.op.isnan and relax.op.where to handle NaN values, which is a robust solution. The addition of a dedicated test case for NaN inputs is also a great improvement. I have one minor suggestion to make the creation of the NaN constant more direct and efficient.

[gemini-code-assist]

For creating a NaN scalar constant, it's more direct and readable to use relax.const(float('nan'), dtype=dtype) instead of dividing zero by zero. This also avoids creating intermediate Constant and divide operator nodes in the graph, making the graph construction slightly more efficient.

Suggested change

	nan_val = relax.op.divide(relax.const(0.0, dtype=dtype), relax.const(0.0, dtype=dtype))
	nan_val = relax.const(float("nan"), dtype=dtype)