[BUGFIX][NON_CUDA] Fix failing introduced by #20061 when import bitsandbytes.py

[xuechendi]

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.

Purpose

This PR is to fixed an exception introduce by #20061
#20061 will do import in bitsandbytes.py

from vllm.model_executor.layers.fused_moe import fused_experts 

However, in vllm/model_executor/layer.fused_moe/init.py
fused_moe is imported under condition

if HAS_TRITON: # import to register the custom ops import vllm.model_executor.layers.fused_moe.fused_marlin_moe # noqa import vllm.model_executor.layers.fused_moe.fused_moe # noqa 

For any platform doesn't enable Triton, will hit error as below

For HW doesn't has_triton enabled, when import bitsandbytes, it will trigger error as below:

vllm-gaudi/tests/models/language/generation/test_common.py:52: in launch_lm_eval llm = VLLM(**model_args) ^^^^^^^^^^^^^^^^^^ /usr/local/lib/python3.12/dist-packages/lm_eval/models/vllm_causallms.py:177: in __init__ self.model = LLM(**self.model_args) ^^^^^^^^^^^^^^^^^^^^^^ /usr/local/lib/python3.12/dist-packages/vllm-0.9.2rc2.dev178+g5f0af36af.gaudi000-py3.12.egg/vllm/entrypoints/llm.py:274: in __init__ self.llm_engine = LLMEngine.from_engine_args( /usr/local/lib/python3.12/dist-packages/vllm-0.9.2rc2.dev178+g5f0af36af.gaudi000-py3.12.egg/vllm/v1/engine/llm_engine.py:142: in from_engine_args vllm_config = engine_args.create_engine_config(usage_context) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ /usr/local/lib/python3.12/dist-packages/vllm-0.9.2rc2.dev178+g5f0af36af.gaudi000-py3.12.egg/vllm/engine/arg_utils.py:1104: in create_engine_config model_config = self.create_model_config() ^^^^^^^^^^^^^^^^^^^^^^^^^^ /usr/local/lib/python3.12/dist-packages/vllm-0.9.2rc2.dev178+g5f0af36af.gaudi000-py3.12.egg/vllm/engine/arg_utils.py:976: in create_model_config return ModelConfig( /usr/local/lib/python3.12/dist-packages/vllm-0.9.2rc2.dev178+g5f0af36af.gaudi000-py3.12.egg/vllm/config.py:631: in __post_init__ self._verify_quantization() /usr/local/lib/python3.12/dist-packages/vllm-0.9.2rc2.dev178+g5f0af36af.gaudi000-py3.12.egg/vllm/config.py:935: in _verify_quantization method = me_quant.get_quantization_config(name) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ /usr/local/lib/python3.12/dist-packages/vllm-0.9.2rc2.dev178+g5f0af36af.gaudi000-py3.12.egg/vllm/model_executor/layers/quantization/__init__.py:94: in get_quantization_config from .bitsandbytes import BitsAndBytesConfig _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ # SPDX-License-Identifier: Apache-2.0 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project from typing import Any, Callable, Optional, Union import torch > from vllm.model_executor.layers.fused_moe import fused_experts E ImportError: cannot import name 'fused_experts' from 'vllm.model_executor.layers.fused_moe' (/usr/local/lib/python3.12/dist-packages/vllm-0.9.2rc2.dev178+g5f0af36af.gaudi000-py3.12.egg/vllm/model_executor/layers/fused_moe/__init__.py) 

Test Plan

Test Result

(Optional) Documentation Update

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

[gemini-code-assist]

Summary of Changes

Hello @xuechendi, 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 import error that was preventing the bitsandbytes quantization layer from being properly initialized on systems without Triton support. By adjusting the import logic to be conditional on the presence of Triton, the change ensures broader compatibility and prevents runtime failures during VLLM setup.

Highlights

• Bug Fix: This pull request resolves an ImportError that occurred when vllm.model_executor.layers.quantization.bitsandbytes was imported on systems where Triton is not available. This issue was introduced by a previous change (#20061) which unconditionally imported a Triton-dependent module.
• Conditional Import Logic: The import of fused_experts within vllm/model_executor/layers/quantization/bitsandbytes.py has been made conditional on the HAS_TRITON flag. This change aligns the import behavior with the conditional availability of Triton-dependent modules, preventing crashes on non-Triton environments.

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. ↩

[xuechendi]

@jeejeelee @Isotr0py , may you help to review this PR, suggest a quick fix in bitsandbytes to avoid failing due to incorrect import.
Thanks so much

[gemini-code-assist]

Code Review

This pull request correctly fixes an ImportError on non-Triton platforms by making the import of fused_experts conditional. My feedback includes a suggestion to define fused_experts as None when Triton is not available to make the code more robust against potential NameError exceptions and to allow for more explicit error handling.

[Isotr0py]

I think #20822 should fix this issue as well?

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @xuechendi.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[xuechendi]

Fixed by other PR, close this one