bug: flaky test case Functionality test for helm chart / Multiple-Models

[gaocegege]

Describe the bug

In the PR #119 the test The test Functionality test for helm chart / Multiple-Models (pull_request) is flaky. Sometimes it fails in helm uninstall, and sometimes fails in helm install.

To Reproduce

N/A

Expected behavior

No response

Additional context

No response

[Shaoting-Feng]

I reviewed the GitHub Actions records and believe the issue was caused by the workflow server being occupied at that time. We will continue monitoring the workflow to determine whether there is a bug and, if so, identify the exact error message.

[ApostaC]

Currently, the functionality test is running on a self-hosted GPU runner we host.

I think we should move to a cpu-only vLLM image for functionality testing so that we can use github's runner.