[bugfix][AWQ] make compute_layer_means not modify weight

[HDCharles]

Summary

_compute_layer_means was making in-place changes to the weights which made the first iteration (until the weights are refreshed from the cached versions) useless

flow was

1. cache weight
2. compute mean (modify weight)
3. scale weight
4. calculate loss
5. restore weight, goto 3

so the first run of 3&4 were gibberish

note this bug has been around for a while and we hadn't detected it because throwing away a single iteration was fine as long as the rest worked

TEST PLAN: (new unit test)
python tests/llmcompressor/modifiers/awq/test_base.py

[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 @HDCharles, 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 a critical bug within the AWQ (Activation-aware Weight Quantization) module, specifically in the _compute_layer_means function. Previously, this function was inadvertently performing in-place modifications on the model's weight tensors, which could result in erroneous computations, particularly during the initial processing pass. The fix introduces a cloning mechanism for the weight tensor before any operations, thereby safeguarding the original weights and ensuring the integrity and consistency of the quantization process.

Highlights

• Bug Fix: Resolved a bug in the _compute_layer_means function where weight tensors were being modified in-place, leading to incorrect calculations during the first iteration.
• Weight Tensor Handling: Ensured that _compute_layer_means now operates on a cloned copy of the weight tensor, preserving the original weights and guaranteeing consistent behavior across all iterations.

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 fixes a bug in _compute_layer_means where layer weights were being modified in-place. By cloning the weight tensor before modification, the change correctly prevents side effects on the original model weights. This is a good and direct fix for the issue, improving the correctness of the implementation. The change is minimal and well-justified.