Update dependency bounds for llmcompressor 0.9 release #2086

dhuangnm · 2025-12-02T18:26:39Z

SUMMARY:
Update dependency bounds based on their latest releases.

TEST PLAN:
All nightly tests.

Run with upper bounds: https://github.com/neuralmagic/llm-compressor-testing/actions/runs/19864870462
Run with lower bounds: https://github.com/neuralmagic/llm-compressor-testing/actions/runs/19867046299

github-actions · 2025-12-02T18:26:50Z

👋 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 · 2025-12-02T18:26:53Z

Summary of Changes

Hello @dhuangnm, 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 focuses on updating the dependency version constraints within the setup.py file for the llmcompressor project. The changes primarily involve raising the upper bounds for several key libraries like numpy, torch, transformers, datasets, and pillow, allowing the project to utilize more recent versions of these packages. Additionally, the accelerate dependency's range has been refined to ensure compatibility with auto-round 0.9.1. These updates are crucial for maintaining compatibility with the latest ecosystem and preparing for the llmcompressor 0.9 release.

Highlights

Dependency Upper Bounds Updated: Increased the maximum allowed versions for numpy, torch, transformers, datasets, and pillow to incorporate newer releases, ensuring compatibility with the latest ecosystem.
Accelerate Dependency Refined: Adjusted the accelerate dependency's lower bound to >=1.10.0 and upper bound to <=1.12.0 for release builds, specifically to meet the requirements of auto-round 0.9.1.

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.

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. ↩

Signed-off-by: Dan Huang <[email protected]>

gemini-code-assist

Code Review

This pull request updates the dependency bounds for several packages. The changes are mostly straightforward version bumps. However, I've identified a potential dependency issue with accelerate in non-release builds, which could lead to conflicts with auto-round. I've also pointed out a redundant conditional for the auto-round dependency that can be simplified. My suggestions aim to ensure dependency consistency across different build types and improve code clarity.

gemini-code-assist · 2025-12-02T18:27:49Z

setup.py

-            "accelerate>=1.6.0,<=1.10.1"
+            "accelerate>=1.10.0,<=1.12.0"
            if BUILD_TYPE == "release"
            else "accelerate>=1.6.0"


The accelerate dependency for non-release builds is set to >=1.6.0. However, the comment on line 131 states that auto-round==0.9.1 requires accelerate >= 1.10.0. Since auto-round is a dependency for all build types, the lower bound for accelerate should be 1.10.0 for non-release builds as well to avoid potential dependency conflicts.

Suggested change

else "accelerate>=1.6.0"

else "accelerate>=1.10.0"

gemini-code-assist · 2025-12-02T18:27:50Z

setup.py

            else "compressed-tensors>=0.12.3a2"
        ),
-        ("auto-round==0.9.1"),
+        ("auto-round==0.9.1" if BUILD_TYPE == "release" else "auto-round==0.9.1"),


The conditional logic for the auto-round dependency is redundant as it specifies the same version 0.9.1 for both release and non-release builds. This can be simplified to a static dependency for better readability and maintainability.

Suggested change

("auto-round==0.9.1" if BUILD_TYPE == "release" else "auto-round==0.9.1"),

("auto-round==0.9.1"),

I think we may want to update this to be its own install tag, similar to dev

e. g uv pip install -e .[dev,autoround]

Updated, I'll file another PR for the workflows to use the new [autoround] install tag.

https://github.com/neuralmagic/llm-compressor-testing/pull/183 to update the workflows accordingly.

dsikka

Thanks!

Just an FYI - I think we want to pin the transformers version, regardless of release. This is for the time being while we ensure we can support the v5 release that will be out soon.

Signed-off-by: Dan Huang <[email protected]>

dhuangnm · 2025-12-02T20:35:50Z

Thanks!

Just an FYI - I think we want to pin the transformers version, regardless of release. This is for the time being while we ensure we can support the v5 release that will be out soon.

Got it, updated the upper bound for transformers to 5.0.0 although currently its latest is 5.0.0.rc0 on the PyPi.

Signed-off-by: Dan Huang <[email protected]>

dsikka · 2025-12-02T21:03:39Z

Thanks!
Just an FYI - I think we want to pin the transformers version, regardless of release. This is for the time being while we ensure we can support the v5 release that will be out soon.

Got it, updated the upper bound for transformers to 5.0.0 although currently its latest is 5.0.0.rc0 on the PyPi.

Oh sorry, I meant the upper bound should be the current stable release

dsikka · 2025-12-02T21:05:22Z

setup.py

-        ("auto-round==0.9.1"),
    ],
    extras_require={
+        "autoround": [


Thanks! We will also need to change the transformers test workflow to run autoround tests as they require this install:

llm-compressor/.github/workflows/test-check-transformers.yaml

Line 104 in 6284e1b

- name: Running AutoRound Tests

If we think this will be a bigger change, we can make the autoround optional install change in a followup

Good catch, updated the workflow to install "autoround" dependency as well.

Should be a separate job, no?
Sorry, I think my request is becoming larger than we expected. We can revert to install autoround how you were previously (i.e not using the optional install) and I'll add a follow-up to fix this

Signed-off-by: Dan Huang <[email protected]>

dhuangnm · 2025-12-02T21:36:09Z

Thanks!
Just an FYI - I think we want to pin the transformers version, regardless of release. This is for the time being while we ensure we can support the v5 release that will be out soon.

Got it, updated the upper bound for transformers to 5.0.0 although currently its latest is 5.0.0.rc0 on the PyPi.

Oh sorry, I meant the upper bound should be the current stable release

Gotcha, updated.

dhuangnm requested review from dbarbuzzi, dsikka, fynnsu and kylesayrs December 2, 2025 18:26

dhuangnm added 2 commits December 2, 2025 13:27

update upper bounds for 0.9 release

fe7b04a

Signed-off-by: Dan Huang <[email protected]>

update accelerate based on auto-round requirement

09baec8

Signed-off-by: Dan Huang <[email protected]>

gemini-code-assist bot reviewed Dec 2, 2025

View reviewed changes

dhuangnm force-pushed the upper branch from c0b360f to 09baec8 Compare December 2, 2025 18:28

fynnsu previously approved these changes Dec 2, 2025

View reviewed changes

dsikka reviewed Dec 2, 2025

View reviewed changes

update transformers and auto-round

19fb2b2

Signed-off-by: Dan Huang <[email protected]>

dhuangnm dismissed fynnsu’s stale review via 19fb2b2 December 2, 2025 20:32

fix format

c79b83c

Signed-off-by: Dan Huang <[email protected]>

dsikka reviewed Dec 2, 2025

View reviewed changes

dhuangnm added 2 commits December 2, 2025 16:32

update transformers

be3a99f

Signed-off-by: Dan Huang <[email protected]>

update transformers test workflow

71b5474

Signed-off-by: Dan Huang <[email protected]>

	("auto-round==0.9.1" if BUILD_TYPE == "release" else "auto-round==0.9.1"),
	("auto-round==0.9.1"),

Update dependency bounds for llmcompressor 0.9 release #2086

Are you sure you want to change the base?

Update dependency bounds for llmcompressor 0.9 release #2086

Uh oh!

Conversation

dhuangnm commented Dec 2, 2025

Uh oh!

github-actions bot commented Dec 2, 2025

Uh oh!

gemini-code-assist bot commented Dec 2, 2025

Summary of Changes

Highlights

Footnotes

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Dec 2, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Dec 2, 2025

Choose a reason for hiding this comment

Uh oh!

dsikka Dec 2, 2025

Choose a reason for hiding this comment

Uh oh!

dhuangnm Dec 2, 2025

Choose a reason for hiding this comment

Uh oh!

dhuangnm Dec 2, 2025

Choose a reason for hiding this comment

Uh oh!

dsikka left a comment

Choose a reason for hiding this comment

Uh oh!

dhuangnm commented Dec 2, 2025

Uh oh!

dsikka commented Dec 2, 2025

Uh oh!

dsikka Dec 2, 2025

Choose a reason for hiding this comment

Uh oh!

dhuangnm Dec 2, 2025

Choose a reason for hiding this comment

Uh oh!

dsikka Dec 2, 2025

Choose a reason for hiding this comment

Uh oh!

dhuangnm commented Dec 2, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants