[bugfix] fix mllm megatron gkd sft_alpha

[hjh0119]

#8366

[gemini-code-assist]

Summary of Changes

Hello, 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 resolves a bug in the MLLM Megatron GKD trainer related to the sft_alpha loss calculation. The change ensures that the compute_language_model_loss method is consistently called on the correct underlying language model component, improving the robustness of the training process.

Highlights

• Loss Calculation Fix: Corrected the method for calculating per_token_loss within the GKD trainer's loss_func when sft_alpha is active.
• Model Component Access: Implemented a check to ensure that compute_language_model_loss is invoked on the specific language_model attribute of the unwrapped model, if it exists, to correctly handle different model structures.

🧠 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.

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.

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 addresses a bug in the GKD trainer's SFT loss calculation for multimodal models. The change correctly retrieves the underlying language model, ensuring compatibility with MLLM architectures that have a wrapper structure. The fix is sound. I have provided one suggestion to improve code conciseness and readability.

[gemini-code-assist]

For improved readability and conciseness, you can use getattr with a default value. This avoids reassigning the model variable and makes the logic for obtaining the correct model more direct and idiomatic.

            language_model = getattr(self.unwrapped_models[0], 'language_model', self.unwrapped_models[0])
            per_token_loss = language_model.compute_language_model_loss(labels, logits_sbv)