[Relax] Clean up scatter_elements unknown dtype handling

[guan404ming]

Why

• LOG(WARNING) is the standard and correct approach throughout the TVM codebase
• The existing pattern is used consistently in all relax ops (see test_op_manipulate.py, index.cc, etc.)
• Added test coverage for previously untested scenarios

[gemini-code-assist]

Summary of Changes

Hello @guan404ming, 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 refines the scatter_elements operator's behavior when encountering unknown data types. It standardizes the warning mechanism to LOG(WARNING) and removes outdated TODO comments, aligning with existing TVM codebase practices. Additionally, it enhances test coverage to ensure proper handling of previously untested scenarios involving unknown data and update tensor data types.

Highlights

• Standardized Warning Mechanism: The handling of unknown data types in the scatter_elements operator now consistently uses LOG(WARNING), aligning with standard practices across the TVM codebase.
• Removed TODO Comments: Outdated TODO comments related to warning reporting for unknown data types in InferStructInfoScatterElements have been removed.
• Enhanced Test Coverage: New test cases have been added for scatter_elements to specifically cover scenarios where the data or updates tensors have unknown data types, ensuring more robust behavior.

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 cleans up the handling of unknown dtypes in scatter_elements by removing some TODO comments and adds test coverage for these scenarios. The changes are good, but I've found a few areas for improvement.
In src/relax/op/tensor/manipulate.cc, I've identified a potential bug where the wrong variable is used to check the dtype of updates, and a minor typo in a log message.
In tests/python/relax/test_op_manipulate.py, while the added tests are valuable, I've suggested adding another test case for when indices has an unknown dtype to ensure full coverage of the warning paths.

[gemini-code-assist]

While you're cleaning up this section, I noticed a potential bug in the usage of the diag_dtype lambda. On line 2406 (in the full file), it seems data_sinfo is used for both "data" and "updates", but it should probably be updates_sinfo for "updates". This looks like a copy-paste error. It should be diag_dtype(updates_sinfo, "updates"); to correctly check the updates tensor.

[gemini-code-assist]

While removing this TODO, I noticed a small typo in the following log message on line 2419. It says "indice" but it should probably be "indices" to be consistent with the variable name indices_sinfo and other similar log messages in the codebase (e.g., in InferStructInfoScatterND).

[gemini-code-assist]

These are good additions for testing unknown dtypes for data and updates. To improve test coverage further, could you also add a test case for when indices has an unknown dtype? This would cover the warning log path for indices_sinfo->IsUnknownDtype() in InferStructInfoScatterElements.

    # Test with unknown dtype for data
    d_unknown = relax.Var("data", R.Tensor((4, 4)))
    _check_inference(
        bb,
        relax.op.scatter_elements(d_unknown, i0, u0, 0, "updates"),
        relax.TensorStructInfo((4, 4), dtype=""),
    )
    # Test with unknown dtype for updates
    u_unknown = relax.Var("updates", R.Tensor((2, 2)))
    _check_inference(
        bb,
        relax.op.scatter_elements(d0, i0, u_unknown, 0, "updates"),
        relax.TensorStructInfo((4, 4), dtype="float32"),
    )
    # Test with unknown dtype for indices
    i_unknown = relax.Var("indices", R.Tensor((2, 2)))
    _check_inference(
        bb,
        relax.op.scatter_elements(d0, i_unknown, u0, 0, "updates"),
        relax.TensorStructInfo((4, 4), dtype="float32"),
    )

[guan404ming]

cc @tlopex @mshr-h

[tlopex]

LGTM! Thanks