[bugfix] fix apply_wd_to_qk_layernorm

[Jintao-Huang]

#7978

[gemini-code-assist]

Summary of Changes

Hello @Jintao-Huang, 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 primarily addresses a bug related to the application of weight decay (implied by the title) in multimodal models. It refines the logic within the _get_param_groups function to correctly handle multimodal-specific parameters, such as visual module learning rates, by conditionally accessing multimodal components. This prevents errors when the model is not multimodal. Additionally, a minor adjustment was made to a shell script, removing nohup from a swift sft command.

Highlights

• Multimodal Model Logic Refinement: Improved handling of multimodal model-specific parameters and components in the _get_param_groups function, ensuring that visual module access and learning rate assertions are only applied when the model is indeed multimodal.
• Script Execution Adjustment: Removed the nohup command from a swift sft execution script, which might affect how the process runs in the background.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

Changelog

• examples/models/qwen3_next/transformers.sh
  • The nohup command was removed from the swift sft command line, changing its execution behavior (e.g., no longer running detached from the terminal).
• swift/megatron/trainers/base.py
  • A local variable is_multimodal was introduced to explicitly check the model's multimodal status.
  • The assertion message for vit_lr and aligner_lr was improved to clearly state that these learning rates are only supported for multimodal models.
  • Access to the visual attribute of the model chunk was made conditional on is_multimodal, preventing potential errors when processing non-multimodal models.

Activity

• No human activity (comments, reviews) has been recorded on this pull request yet.

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 introduces a bugfix to prevent an AttributeError in _get_param_groups when running with non-multimodal models, which could be triggered when apply_wd_to_qk_layernorm is enabled. The fix correctly adds a check for is_multimodal before accessing the visual attribute of the model. The changes also include a minor refactoring to cache the is_multimodal flag and improve an assertion message, which enhances code readability and maintainability. The change in the example script is a runtime adjustment. Overall, the changes are correct and improve the code's robustness.