[Bugfix] qwen3-vl-235b-w8a8 load weight ERROR when start service

[Levi-JQ]

What this PR does / why we need it?

fix qwen3-v-w8a8 load weight ERROR when start service

Does this PR introduce any user-facing change?

How was this patch tested?

• vLLM version: v0.11.0
• vLLM main: vllm-project/vllm@2918c1b

[github-actions]

👋 Hi! Thank you for contributing to the vLLM Ascend project. The following points will speed up your PR merge:‌‌

• A PR should do only one thing, smaller PRs enable faster reviews.
• Every PR should include unit tests and end-to-end tests ‌to ensure it works and is not broken by other future PRs.
• Write the commit message by fulfilling the PR description to help reviewer and future developers understand.

If CI fails, you can run linting and testing checks locally according Contributing and Testing.

[gemini-code-assist]

Code Review

This pull request introduces a bugfix for weight loading in quantized qwen3_vl_moe models. It correctly adds prefix mapping logic to find quantization parameters and defines packed modules for this model type. My review focuses on improving the maintainability of the prefix mapping logic. I've suggested a change to avoid duplicating configuration data, which will make the code clearer and less prone to future errors.

[gemini-code-assist]

The current implementation for mapping vLLM prefixes to HuggingFace prefixes for qwen3_vl_moe duplicates configuration by defining an inverted mapping dictionary. This is confusing and error-prone if the original mapping changes.

To improve maintainability and clarity, I suggest using the same dictionary structure as in the model definition (HF-to-vLLM) and then using the _reverse_map_name method to perform the required vLLM-to-HF conversion. This makes the code's intent clearer and avoids maintaining two separate, inverted dictionaries.

Suggested change

	if model_type == "qwen3_vl_moe":
	hf_to_vllm_mapper = WeightsMapper(
	orig_to_new_prefix={
	"visual.": "model.visual.",
	"language_model.lm_head.": "lm_head.",
	"language_model.model.": "model.language_model.",
	})
	prefix = hf_to_vllm_mapper._map_name(prefix)
	if model_type == "qwen3_vl_moe":
	hf_to_vllm_mapper = WeightsMapper(
	orig_to_new_prefix={
	"model.visual.": "visual.",
	"lm_head.": "language_model.lm_head.",
	"model.language_model.": "language_model.model.",
	})
	prefix = hf_to_vllm_mapper._reverse_map_name(prefix)