[Test] Add initial multi modal cases of Qwen2.5-VL-7B-Instruct for disaggregated encoder

[yenuo26]

What this PR does / why we need it?

This PR adds disaggregated encoder tests for Qwen2.5-VL-7B-Instruct

Does this PR introduce any user-facing change?

No

How was this patch tested?

by running the test
by running ci

Test Result

test case:
local:

ci:

examples:

• vLLM version: release/v0.12.0

[gemini-code-assist]

Code Review

This PR adds a disaggregated encoder proxy and corresponding e2e tests for Qwen2.5-VL. The changes are well-structured, introducing a proxy server, test helpers in conftest.py, and a new test case. The implementation is mostly solid, but I've found a few issues related to correctness and best practices. Specifically, the health check endpoint in the proxy has a bug, a test helper function uses blocking I/O in an async context, another helper class is not robust to its declared input types, and the test case construction of JSON strings could be improved for safety and readability. My detailed comments and suggestions are below.

[gemini-code-assist]

The __init__ method of DisaggEpdProxy accepts proxy_args as Union[list[str], str], but it doesn't handle the str case. If a string is passed, _start_disagg_proxy will incorrectly unpack it character by character, and __aenter__ will fail on self.proxy_args.index("--port"). For robustness and consistency with RemoteOpenAIServer, you should handle the string case by splitting it into a list of arguments using shlex.split.

Suggested change

	def __init__(self,
	proxy_args: Union[list[str], str] = None,
	env_dict: Optional[dict[str, str]] = None) -> None:
	self.proxy_args = proxy_args
	self.env_dict = env_dict
	self._proc_list = list()
	def __init__(self,
	proxy_args: Union[list[str], str] = None,
	env_dict: Optional[dict[str, str]] = None) -> None:
	if isinstance(proxy_args, str):
	self.proxy_args = shlex.split(proxy_args)
	else:
	self.proxy_args = proxy_args
	self.env_dict = env_dict
	self._proc_list = list()

[gemini-code-assist]

Constructing JSON strings via concatenation is error-prone and hard to read. It's better to define the structure as a Python dictionary and use json.dumps() to create a valid JSON string. This improves readability and correctness. Please also remember to import json at the top of the file.

        json.dumps({
            "ec_connector_extra_config": {
                "shared_storage_path": SHARED_STORAGE_PATH
            },
            "ec_connector": "ECSharedStorageConnector",
            "ec_role": "ec_producer"
        })

[gemini-code-assist]

Constructing JSON strings via concatenation is error-prone and hard to read. It's better to define the structure as a Python dictionary and use json.dumps() to create a valid JSON string. This improves readability and correctness. Please also remember to import json at the top of the file.

        json.dumps({
            "ec_connector_extra_config": {
                "shared_storage_path": SHARED_STORAGE_PATH
            },
            "ec_connector": "ECSharedStorageConnector",
            "ec_role": "ec_consumer"
        })

[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.

[MengqingCao]

Is there any difference between this example and https://github.com/vllm-project/vllm/blob/main/examples/online_serving/disaggregated_encoder/disagg_epd_proxy.py?

[MengqingCao]

Suggested change

self.env_dict['VLLM_USE_V1'] = "1"

[wangxiyuan]

Any progress? If this PR is still alive, please rebase to main and make CI happy, otherwise you can close it. Thanks

[github-actions]

This pull request has conflicts, please resolve those before we can evaluate the pull request.

[yenuo26]

适配vllm新版本

[MrZ20]

Hi! The nighltly test failed; there seem to be some errors that you need to resolve.
https://github.com/vllm-project/vllm-ascend/actions/runs/21801287662/job/62907968683

[yenuo26]

OK，i will check it