[v1] add renderer ut

[hiyouga]

What does this PR do?

Fixes # (issue)

Before submitting

• Did you read the contributor guideline?
• Did you write any new necessary tests?

[gemini-code-assist]

Summary of Changes

Hello @hiyouga, 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 introduces a specialized qwen3_nothink template for Qwen3 models, enhancing their ability to handle complex chat interactions including tool calls and reasoning. It involves a significant refactoring of the message rendering and parsing infrastructure, moving model-specific logic into a plugin architecture while simplifying the generic ChatML renderer. The changes also include improvements to type safety in loss functions, a more robust PyTorch version check, and updated data structures for tool arguments, all supported by new unit tests to ensure correctness.

Highlights

• New Qwen3 Template: A new qwen3_nothink template has been introduced for Qwen3 models, providing specific logic for message rendering and parsing, including handling of tool calls and reasoning tags.
• Refactored Rendering Logic: The detailed message rendering and parsing logic, previously in src/llamafactory/v1/core/utils/rendering.py, has been moved and implemented as a plugin in src/llamafactory/v1/plugins/model_plugins/rendering.py for the new qwen3_nothink template. This includes the _update_model_input helper function and complex parsing for various message roles and content types.
• Simplified ChatML Rendering: The generic render_chatml_messages and parse_chatml_message functions in src/llamafactory/v1/core/utils/rendering.py have been significantly simplified, now focusing on basic text messages and removing the complex tool call and reasoning parsing.
• Improved Type Hinting for Loss Functions: Explicit type hints have been added to parameters like outputs, labels, num_items_in_batch, source, target, and alpha in the dft_loss_func, _dft_cross_entropy, eaft_loss_func, and _eaft_cross_entropy functions for better code clarity and maintainability.
• Enhanced PyTorch Version Check: The utility for checking PyTorch versions has been updated, replacing _get_package_version and packaging.version with a new is_torch_version_greater_than function, specifically used for the Conv3D warning.
• Updated ToolCall Type Definition: The arguments field within the ToolCall TypedDict in src/llamafactory/v1/utils/types.py has been changed from str to dict[str, Any], allowing for a more structured and flexible representation of tool arguments.
• New Unit Tests for Qwen3 Template: Comprehensive unit tests have been added in tests_v1/core/utils/test_rendering.py to validate the correct functioning of the new qwen3_nothink template's rendering and parsing logic, including scenarios involving tool calls.
• Default Template Change: The default template in src/llamafactory/v1/config/model_args.py has been updated from chatml to qwen3_nothink.

🧠 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.

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 refactors the rendering logic by moving model-specific rendering for Qwen models into a plugin, which improves modularity. It also adds comprehensive unit tests for this new renderer, ensuring its correctness against the reference implementation from transformers. The changes also include some code cleanup, such as removing unused imports and adding type hints, which enhances code quality. Overall, this is a good improvement to the codebase. I have a few suggestions to improve code readability and test consistency.