Add Gemma3 Conversion script to port weights from HF directly

[laxmareddyp]

Description of the change

Reference

Colab Notebook

Checklist

• 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.

[gemini-code-assist]

Summary of Changes

Hello @laxmareddyp, 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 expands KerasHub's capabilities by adding comprehensive support for Gemma3 models. It provides the necessary infrastructure to convert Gemma3 model architectures and weights from Hugging Face, integrates this conversion into the existing preset loading system, and offers a dedicated tool for converting and validating Gemma3 checkpoints, ensuring seamless interoperability and ease of use for Keras developers.

Highlights

• Gemma3 Conversion Script: A new Python script (convert_gemma3.py) has been added to enable the direct porting of Gemma3 model weights and configurations from Hugging Face to KerasHub, supporting both text-only and multi-modal variants.
• Preset Loader Integration: The preset_loader.py utility has been updated to recognize and utilize the new Gemma3 conversion logic, allowing KerasHub to load Gemma3 models directly from Hugging Face presets.
• Checkpoint Conversion Tool: A new command-line tool (convert_gemma3_hf_checkpoints.py) is introduced to facilitate the end-to-end conversion of Gemma3 checkpoints from Hugging Face, including validation of parameter counts, tokenizer outputs, and model logits, before saving them as KerasHub presets.

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 adds a conversion script for Gemma3 models, allowing weights to be ported directly from Hugging Face. The changes include the main conversion logic in convert_gemma3.py, updates to preset_loader.py to recognize Gemma3 models, and a new checkpoint conversion script convert_gemma3_hf_checkpoints.py for validation.

The implementation is well-structured and follows the existing patterns for model conversion in the repository. The validation script is comprehensive, checking for parameter count, tokenizer output, and model logits, which is great for ensuring correctness.

My main feedback is on the convert_gemma3.py file, where some of the weight transformation logic can be refactored into helper functions to improve readability and reduce code duplication. This is a minor point on an otherwise solid contribution.