[Bugfix] Fix typo in Qwen3 Next model executor

[Nepherpitou]

Purpose

There is an issue in Qwen3Next model loading because of typo with paddings.

Sorry, not sure if PR description is sufficient. But bug is obvious and I fixed it by modifying files in .venv. Just consider it will be nice to have here as well.

Test Plan

• Load Qwen3 Next model

Test Result

• Loaded successfully

---

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.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

[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 ask your reviewers to trigger select CI tests on top of fastcheck CI.

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.

If you have any questions, please reach out to us on Slack at https://slack.vllm.ai.

🚀

[gemini-code-assist]

Code Review

This pull request addresses a critical bug in the set_moe_parameters method within the Qwen3Next model executor. The original implementation had an indentation error, causing a check for Mixture-of-Experts (MoE) layers to be performed on every iteration of a loop. This would incorrectly raise a RuntimeError if the first layer of a model was not an MoE layer, effectively preventing many models from loading. The fix correctly moves this check to after the loop, ensuring that all layers are evaluated before concluding that no MoE layers are present. The change is correct and essential for the model's functionality.