[Fix] Fix several issues preventing gemma3n LoRA support.#8776
Conversation
There was a problem hiding this comment.
Summary of Changes
Hello @lifuhuang, 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!
I've implemented several fixes to enable and improve LoRA (Low-Rank Adaptation) support for the Gemma3n model. This includes enhancing the validation of LoRA configurations, integrating the specific module identification and dimension mapping required for Gemma3n's architecture, and addressing known compatibility issues with hybrid SWA memory for these models. The overall goal is to ensure that Gemma3n models can be effectively fine-tuned using LoRA.
Highlights
- Improved LoRA Configuration Validation: I've added a new validation step in the LoRA manager to ensure that
target_modulesare always specified as a list. This helps prevent configuration errors and provides clearer guidance to users. - Enabled LoRA Support for Gemma3n: I've integrated the necessary logic within the Gemma3n model definition to properly identify LoRA-applicable modules and determine their input/output dimensions. This is a critical step for enabling LoRA support for Gemma3n models.
- Addressed Hybrid SWA Memory Compatibility for Gemma3n: I've updated the server arguments to disable hybrid SWA memory for Gemma3n models, similar to Gemma2, due to existing compatibility issues. This ensures stability while these models are in use.
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 in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.
| 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 issue 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 or fill out our survey to provide feedback.
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
-
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. ↩
There was a problem hiding this comment.
Code Review
This pull request introduces fixes to enable LoRA support for gemma3n models. The changes include adding a validation check for LoRA target modules in lora_manager.py, implementing LoRA-specific methods in gemma3n_mm.py, and updating server arguments to handle gemma3n models correctly. My review focuses on improving the newly added get_hidden_dim method in gemma3n_mm.py for better maintainability by adding type hints and removing code duplication. The other changes look good.
Motivation
See #8775
Modifications
Accuracy Test
Benchmark & Profiling
Checklist