[Feature][EPLB] Add eplb support for Qwen2

[lengrongfu]

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)". [Feature]: Support EPLB for More MoE Models, e.g. Qwen 3, Llama 4 #20468
• 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

Test Plan

Test Result

(Optional) Documentation Update

[gemini-code-assist]

Code Review

This pull request adds support for Expert Parallelism Load Balancing (EPLB) to the Qwen2 MoE model. The changes are mostly confined to vllm/model_executor/models/qwen2_moe.py and involve adding EPLB-related configurations, implementing the MixtureOfExperts interface, and updating the weight loading logic to handle expert replicas. The implementation appears to be heavily inspired by the existing deepseek_v2 model, which is a reasonable approach. I've identified one high-severity issue related to an import that creates an unnecessary dependency, which should be addressed to improve code maintainability.

[gemini-code-assist]

To improve modularity and reduce coupling between different model implementations, MixtureOfExperts should be imported directly from its definition file. It is defined as a protocol in vllm.model_executor.models.interfaces. Importing from another model's implementation file (deepseek_v2) creates an unnecessary dependency.

Suggested change

	from vllm.model_executor.models.deepseek_v2 import MixtureOfExperts
	from vllm.model_executor.models.interfaces import MixtureOfExperts

[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.

🚀

[mergify]

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

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