Add MobileNetV5 to KerasHub

[harshaljanjani]

The current design integrates well within the Gemma 3n scope. Ready for reviews!

Numerics Matching (Precision: 1e-5)

cc: @divyashreepathihalli @mattdangerw

Colab Notebook

Numerics Validation Reproducibility Notebook

• I have added all the necessary unit tests for my change.
• I have verified that my change does not break existing code and works with all backends (TensorFlow, JAX, and PyTorch).
• My PR is based on the latest changes of the main branch (if unsure, rebase the code).
• I have followed the Keras Hub Model contribution guidelines in making these changes.
• I have followed the Keras Hub API design guidelines in making these changes.
• I have signed the Contributor License Agreement.

Closes #2401.

[harshaljanjani]

Let's start the review rounds now that the model numerics match within a 1e-5 tolerance and the design integrates with Gemma 3n!!
Note: The JAX-GPU times out, but the MobileNetV5 tests also pass on JAX-GPU.
No errors are related to MobileNetV5.

[divyashreepathihalli]

/gemini review

[gemini-code-assist]

Code Review

This pull request introduces MobileNetV5 to KerasHub, following the established modular structure with a backbone, classifier, preprocessor, tests, and a checkpoint conversion script. The implementation is comprehensive and well-organized. My primary feedback revolves around ensuring model serializability. Several new custom layers are missing the get_config method, which is critical for saving and loading models. I have detailed these instances in my comments. Additionally, I've identified a couple of latent bugs in the MobileNetV5Builder concerning the handling of the 'ir' block type, which, while not affecting current presets, should be addressed for future robustness. Overall, this is a solid contribution, and addressing the serialization issues is the main priority.

[sachinprasadhs]

Thank you for the contribution, I have added my comment.
Can you add the test smallest preset test, you can mark as skip test for now.
Also, add on the fly conversion from hf here https://github.com/keras-team/keras-hub/tree/master/keras_hub/src/utils/timm

[harshaljanjani]

@sachinprasadhs Resolved all comments; thanks!

Can you add the smallest preset test? You can mark it as a skipped test for now.
Also, please add on-the-fly conversion from HF here: https://github.com/keras-team/keras-hub/tree/master/keras_hub/src/utils/timm

Since only one preset is currently exposed for MobileNetV5, are these two additions really needed? It might make sense to keep them if other presets are later added by TIMM / PyTorch Image Models, but for now, I’m not too sure; since all presets == the smallest preset, and there’s only one preset to load from HF, which we’re already providing with our Kaggle preset.

Would love your suggestions!

[sachinprasadhs]

@sachinprasadhs Resolved all comments; thanks!

Can you add the smallest preset test? You can mark it as a skipped test for now.
Also, please add on-the-fly conversion from HF here: https://github.com/keras-team/keras-hub/tree/master/keras_hub/src/utils/timm

Since only one preset is currently exposed for MobileNetV5, are these two additions really needed? It might make sense to keep them if other presets are later added by TIMM / PyTorch Image Models, but for now, I’m not too sure; since all presets == the smallest preset, and there’s only one preset to load from HF, which we’re already providing with our Kaggle preset.

Would love your suggestions!

Usually for the models ported from HF, we require the on the fly conversion as well so that the model can be loaded from Kaggle as well as HF.

[harshaljanjani]

Usually for the models ported from HF, we require the on the fly conversion as well so that the model can be loaded from Kaggle as well as HF.

Makes sense, will cover the same!
Edit: Done.

[harshaljanjani]

None of the test failures are related to MobileNetV5.

TensorFlow (GPU): GPT-2

JAX (GPU): Timeout

[harshaljanjani]

None of the test failures are related to MobileNetV5.
TensorFlow (GPU): GPT-2
...
JAX (GPU): Timeout
...

The same still holds true.

[sachinprasadhs]

Thanks, overall it looks good to me.
If this model is encoder only weights, not sure if we have to expose MobileNetV5ImageClassifier, can you attach the demo colab to demonstrate this implementation.

[harshaljanjani]

If this model is encoder only weights, not sure if we have to expose MobileNetV5ImageClassifier, can you attach the demo colab to demonstrate this implementation.

As far as I understand, exposing the complete task model allows us to bundle the ported preprocessing with it as well. Would love to know your thoughts, though.
Anyway, it’s completely opt-in, and in terms of code, it’s no more than 200 lines in this PR; the backbone represents the major effort here.
I’ll add the example to the existing demo Colab shortly.
EDIT: Covered here within the existing Colab.