[TEST]Add initial multi modal cases of Qwen2.5-VL-32B-Instruct for nightly test

[yenuo26]

What this PR does / why we need it?

This PR adds the initial multi modal model for nightly test, including 2 cases for Qwen2.5-vl-32b acc/perf test on A3, we need test them daily.

Does this PR introduce any user-facing change?

No

How was this patch tested?

by running the test

vLLM version: v0.11.0rc3
vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

• vLLM version: v0.11.0rc3
• vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

[gemini-code-assist]

Code Review

This pull request adds a new nightly end-to-end test for the Qwen/Qwen2.5-VL-32B-Instruct multi-modal model and includes a fix in aisbench.py for configuration generation. The new test is comprehensive, covering an API check, a multi-modal request, and benchmarks. My review focuses on improving the correctness of the new test. I've suggested using the chat completions API, which is more appropriate for the model under test, to make the test more robust.

[gemini-code-assist]

The model Qwen/Qwen2.5-VL-32B-Instruct is an instruction-tuned chat model. For consistency with other parts of the test (like send_image_request and the aisbench cases which use chat endpoints) and to follow best practices, the initial smoke test should use the chat completions API (client.chat.completions.create) instead of the legacy completions API (client.completions.create). This ensures the correct API is used for the model type and makes the test more robust and maintainable.

Suggested change

	batch = await client.completions.create(
	model=model,
	prompt=prompts,
	**request_keyword_args,
	)
	choices: list[openai.types.CompletionChoice] = batch.choices
	assert choices[0].text, "empty response"
	chat_response = await client.chat.completions.create(
	model=model,
	messages=[{
	"role": "user",
	"content": prompts[0]
	}],
	**request_keyword_args,
	)
	choices = chat_response.choices
	assert choices[0].message.content, "empty response"

[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.

[jiangyunfan1]

CI passed https://github.com/vllm-project/vllm-ascend/actions/runs/18772453760/job/53559572638?pr=3707