Tiny adjust CI testcases

[hnyls2002]

• Also assert eval result when not in CI
• Enable multithread loading to speed up.

[gemini-code-assist]

Summary of Changes

Hello @hnyls2002, 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 significantly expands the model ecosystem by integrating MistralLarge3 and Pixtral models, enhancing the framework's capability to handle diverse architectures. It also brings substantial improvements to FP8 quantization for MoE layers, introducing block-level granularity for better performance and precision. Additionally, it includes refinements to CI testing for DeepseekV3 and a more robust model information retrieval mechanism in the router.

Highlights

• New Model Support: Added comprehensive support for MistralLarge3 and Pixtral models, including their specific architectures, configurations, and weight loading mechanisms.
• FP8 Quantization Enhancements: Introduced block quantization for FP8 Mixture-of-Experts (MoE) layers, improving efficiency and flexibility in quantization strategies. This includes new validation, re-quantization, and kernel alignment for DeepGEMM and FlashInfer.
• CI Testcase Adjustments: Updated DeepseekV3 CI test cases to assert evaluation results consistently and enabled multithreaded loading with a specified number of threads for faster execution.
• Model Configuration Refactoring: Refactored how Mistral-specific model configurations are parsed and adapted to the internal framework, allowing for better handling of various Mistral model variants (e.g., MoE, vision, Eagle).
• Router Manager Improvement: The router manager's get_model_info endpoint now correctly routes to the default or first available router, resolving a previously unimplemented state.

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 significant enhancements, primarily adding support for new models like MistralLarge3 and Pixtral, and implementing block quantization for FP8. The changes are well-structured and include necessary refactoring to accommodate these new features. My review focuses on ensuring the correctness and maintainability of these additions. I've identified a critical missing import that would cause a runtime error, a redundant variable assignment, an unconventional import path, and an issue with boolean representation in a JSON string in the test files. Overall, the changes are substantial and move the project forward, but the identified issues should be addressed.