Deprecate iter_named_leaf_modules and iter_named_quantizable_modules

[kylesayrs]

Purpose

• Refactor module targeting to be cleaner and easier to maintain
• Support skipping TransformBase modules

Prerequisites

• Deprecate iter_named_leaf_modules and iter_named_quantizable_modules neuralmagic/compressed-tensors#381

Changes

• Remove all uses of iter_named_leaf_modules and iter_named_quantizable_modules
• Make Observer inherit from InternalModule

Testing

• https://github.com/neuralmagic/llm-compressor-testing/actions/runs/16123598340 ✅

[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 primarily refactors the codebase by removing custom module iteration utilities and replacing them with standard PyTorch Module iteration methods. This change aims to simplify module traversal logic and align with PyTorch's native API. Additionally, the Observer base class has been updated to inherit from compressed_tensors.InternalModule, and minor code formatting and logic improvements have been applied.

Highlights

• Code Simplification: Replaced custom module iteration functions (iter_named_leaf_modules, iter_named_quantizable_modules) with standard PyTorch model.modules() or module.named_modules() across the codebase. This streamlines module traversal and aligns with native PyTorch API.
• Base Class Update: The Observer class in src/llmcompressor/observers/base.py now inherits from compressed_tensors.InternalModule instead of torch.nn.Module, indicating a shift towards a more specialized base class within the compressed_tensors library.
• Sparsity Check Logic Refinement: Improved the is_sparse24_bitmask_supported function in src/llmcompressor/transformers/compression/sparsity_metadata_config.py to explicitly skip already quantized modules, preventing redundant checks and potentially improving efficiency.

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 is currently in preview and 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 to provide feedback.

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

The changes replace custom module iterators with standard PyTorch iteration. However, the replacements introduce bugs by failing to replicate the original filtering logic, particularly regarding leaf modules and specific quantizable modules. Critical issues in sparsity_metadata_config.py and kv_cache tests need to be addressed.

[brian-dellabetta]

🚀

discussed with @kylesayrs that mapping all iter_named_leaf_modules, regardless of whether include_attn/include_children are False or True, to torch built-in .named_modules() method, should not affect overall behavior

[rahul-tuli]