[Relax] Fix the squeeze operator to behave consistently with torch

[ConvolutedDog]

This commit fixes the squeeze operator to behave consistently with PyTorch
by implementing no-op behavior when squeezing dimensions that are not of size 1.

Previously:
squeeze(x, [1]) on tensor with shape [32, 10, 5] would fail

Now:
squeeze(x, [1]) on tensor with shape [32, 10, 5] returns the original tensor
without modification, matching PyTorch's behavior

This fixes compatibility issues when converting PyTorch models that use
squeeze with dimensions that may not always be 1 during inference."

This work was done in collaboration with guan404ming's commit d87841d.

[gemini-code-assist]

Summary of Changes

Hello @ConvolutedDog, 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 updates the squeeze operator to align its behavior with PyTorch, specifically by making it a no-operation when an attempt is made to remove a dimension that does not have a size of 1. This adjustment eliminates previous errors and enhances the seamless conversion of models from PyTorch, ensuring robust handling of tensor manipulations.

Highlights

• PyTorch Consistency: The squeeze operator now behaves consistently with PyTorch, performing a no-op when attempting to squeeze dimensions not of size 1.
• Error Prevention: Previously, squeezing a non-unit dimension would cause a failure; this is now prevented.
• Model Compatibility: This change improves compatibility for converting PyTorch models that use squeeze with potentially non-unit dimensions.

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 correctly updates the behavior of the squeeze operator in both TOPI and Relax to align with PyTorch, where attempting to squeeze a dimension not of size 1 is a no-op. The changes in include/tvm/topi/transform.h and src/relax/op/tensor/manipulate.cc correctly implement this logic, and the new test case in tests/python/relax/test_frontend_from_exported_program.py validates it. However, the docstrings for the squeeze function in python/tvm/relax/op/manipulate.py, python/tvm/topi/transform.py, and for SqueezeAttrs in include/tvm/relax/attrs/manipulate.h have not been updated to reflect this new behavior, which could cause confusion. I've added comments to address these documentation inconsistencies.

[ConvolutedDog]

@mshr-h @tlopex Could you please re-run the tests?

[tlopex]

@tvm-bot rerun

[mshr-h]

Thanks. Please include testcase for that in the relax op test.
https://github.com/apache/tvm/blob/main/tests/python/relax/test_op_manipulate.py#L840

[tlopex]

LGTM! Thank you

[mshr-h]

LGTM. Thanks!