Support tpu tests allowing tpu precision for matmul

[wenyi-guo]

support more tpu tests

[gemini-code-assist]

Summary of Changes

Hello @wenyi-guo, 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 significantly improves the robustness of Keras's test suite when running on Tensor Processing Units (TPUs). It achieves this by re-integrating a large number of tests that were previously skipped on TPUs and by introducing a mechanism to define and apply specific numerical tolerances for assertions when executing on TPU hardware. This ensures that Keras functionalities are thoroughly validated against the unique precision characteristics of TPUs, leading to more reliable and performant models on these accelerators.

Highlights

• Expanded TPU Test Coverage: A significant number of tests previously excluded from TPU runs have been re-enabled by drastically reducing the excluded_tpu_tests.txt file, allowing for more comprehensive validation on TPU environments.
• TPU-Specific Assertion Tolerances: Introduced tpu_atol, tpu_rtol, and tpu_decimal parameters to assertAllClose and assertAlmostEqual calls across numerous test files. These parameters allow for adjusted precision checks tailored to the numerical characteristics of TPU computations.
• Dynamic TPU Detection: The TestCase class now includes logic to detect if tests are running on a TPU (for both JAX and TensorFlow backends), enabling the dynamic application of TPU-specific assertion tolerances when appropriate.
• Conditional Test Behavior: Implemented conditional logic in specific tests (e.g., test_argmax_negative_zero, test_argmin_negative_zero, test_spectrogram_error) and explicitly skipped test_eig on JAX+TPU, addressing known differences or limitations in TPU computations.

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 enables a large number of TPU tests by introducing TPU-specific precision tolerances. The core changes in test_case.py to support tpu_atol and tpu_rtol are well-implemented. However, I've identified a critical issue in the TensorFlow TPU detection logic that could break tests in non-TPU environments. Additionally, there's a bug in a numpy test and several instances of extremely loose tolerances that could mask real issues on TPUs. Addressing these points will improve the robustness and reliability of the test suite.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 63.15789% with 14 lines in your changes missing coverage. Please review.
✅ Project coverage is 82.61%. Comparing base (74fba84) to head (63f5d72).
⚠️ Report is 2 commits behind head on master.

Files with missing lines	Patch %	Lines
keras/src/testing/test_case.py	61.11%	10 Missing and 4 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #21887      +/-   ##
==========================================
+ Coverage   82.57%   82.61%   +0.03%     
==========================================
  Files         577      578       +1     
  Lines       59650    59819     +169     
  Branches     9356     9391      +35     
==========================================
+ Hits        49254    49417     +163     
+ Misses       7984     7979       -5     
- Partials     2412     2423      +11     

Flag	Coverage Δ
keras	82.42% <57.89%> (+0.03%)	⬆️
keras-jax	62.25% <39.47%> (-0.61%)	⬇️
keras-numpy	57.44% <31.57%> (-0.08%)	⬇️
keras-openvino	34.32% <15.78%> (-0.01%)	⬇️
keras-tensorflow	64.41% <42.10%> (+0.01%)	⬆️
keras-torch	63.57% <31.57%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[hertschuh]

Thanks for all the changes!

[hertschuh]

Oh weird, what's the error?

[wenyi-guo]

NotImplementedError: MLIR translation rule for primitive 'eig' not found for platform tpu

[hertschuh]

Approving to run the tests.

Removing review to re-trigger TPU tests

[hertschuh]

Approving to re-trigger TPU tests