[megatron] fix qwen3_vl new_special_tokens

[Jintao-Huang]

No description provided.

[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 addresses two specific issues related to model and tokenizer handling within the swift framework, particularly concerning Megatron integration and Qwen3-VL models. It refines the logic for resizing model embedding layers when new special tokens are introduced, preventing issues with dummy models, and corrects the tokenizer object passed during Megatron SFT training initialization. These changes aim to improve the robustness and correctness of model setup and training processes.

Highlights

• Refined Tokenizer Resizing Logic: The condition for resizing the model's embedding layer when new special tokens are added now explicitly excludes dummy models. This prevents unnecessary or incorrect resizing operations when a dummy model is being used, ensuring that the model's vocabulary size is only adjusted for actual, functional models.
• Corrected Megatron SFT Initialization: The argument passed to args.init_model_args has been corrected from self.processor to self.tokenizer. This ensures that the correct tokenizer object is used for initializing model arguments within the Megatron SFT training setup, aligning with the expected input type for this function.

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 two targeted fixes for Megatron training, especially concerning multimodal models like qwen3_vl. The change in swift/llm/model/register.py prevents an attempt to resize token embeddings on a dummy model, which is a correct and necessary safeguard. The modification in swift/megatron/train/sft.py ensures the correct tokenizer object is used to determine vocabulary size, fixing a bug for multimodal setups. Both changes are accurate and enhance the code's stability. The implementation is sound.