fix: Kimi-Linear with AWQ quants

[AlpinDale]

No description provided.

[gemini-code-assist]

Summary of Changes

Hello @AlpinDale, 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 implements a fix for the Kimi-Linear model to ensure proper functionality with AWQ quantization. The changes involve refining the internal module naming conventions by adjusting prefixes during the initialization of specific sub-components, such as shared experts and the Mixture-of-Experts layer. This guarantees that the quantization process accurately identifies and applies to the intended parts of the model, thereby resolving issues related to AWQ quants.

Highlights

• Module Prefixing Correction: Added a specific prefix for 'shared_experts' during initialization within the Kimi-Linear model, ensuring proper identification of these components.
• KimiMoE Prefix Renaming: Updated the prefix for the 'KimiMoE' component from a generic 'mlp' to a more descriptive 'block_sparse_moe', enhancing clarity and consistency.
• AWQ Quantization Fix: These prefix adjustments are crucial for correctly applying AWQ (Activation-aware Weight Quantization) to the Kimi-Linear model, resolving potential issues with quantization targeting.

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 small but important fixes for loading Kimi-Linear models, particularly when using AWQ quantization. The changes correct the prefixes used for naming MoE and shared expert layers. Specifically, the prefix for the MoE block is changed from .mlp to .block_sparse_moe, which aligns with standard naming conventions for such layers. Additionally, a prefix is added for the shared_experts MLP, ensuring that its weights are correctly namespaced. These changes are crucial for correct weight loading and appear to be well-implemented.