[Transform] SpinQuant R4

[kylesayrs]

Purpose

• Support R4 transforms before R3. R3 requires hooking into the attention module, where as R4 does not

Prerequisites

• [Transform] [Quantization] Add transforms to compressed tensors vllm#22486

Testing

• TODO:

[github-actions]

👋 Hi! Thank you for contributing to llm-compressor. Please add the ready label when the PR is ready for review.

Note: This is required to complete the testing suite, please only add the label once the PR is code complete and local testing has been performed.

[gemini-code-assist]

Summary of Changes

Hello @kylesayrs, 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 the implementation of SpinQuant R4 transformations, allowing them to be applied independently. This is a significant step as R4 transformations do not require hooking into the attention module, simplifying their application compared to R3. The changes primarily involve enabling the R4 scheme definition and updating related NotImplementedError messages to reflect this new capability.

Highlights

• SpinQuant R4 Scheme Implementation: The _create_r4_scheme method has been fully implemented to define the transformation scheme for SpinQuant R4. This includes specifying the transform_type, randomize, requires_grad, precision, and the apply arguments with TransformArgs targeting mlp_out for both 'input' and 'weight_input' locations.
• Refined R3 and R4 Scheme Status Messages: The NotImplementedError messages for both R3 and R4 scheme creation methods have been updated. R4's error message was removed as it's now implemented, and R3's message was refined to specifically state that R3 rotations are still a future addition, clearly separating the status of R3 and R4.
• Minor Error Message Refinement: A minor textual change was made to the NotImplementedError message within the validate_not_implemented field validator, updating 'right now' to 'as of now' for unsupported 'randomize' and 'learnable' fields.

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 in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.

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 issue 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 support for SpinQuant R4 transforms, which is a good step forward. The implementation of _create_r4_scheme correctly defines the transform scheme for R4 by targeting the MLP output layers, which aligns with the method described in the SpinQuant paper. The changes are clear and well-contained. I have one minor suggestion to ensure file formatting consistency.

[gemini-code-assist]

It's a good practice to end files with a newline character. This helps prevent issues with file concatenation and some version control systems. This file is missing one.

Suggested change

	)
	)