a cluster with a custom deployment strategy

How to deploy a cluster based on qwen3-14b with tp=2, replica=2, supporting kv cache offloading, and using a relatively new version of vllm? Is there a corresponding yaml demo for reference? Currently, it seems difficult to configure a yaml that suits one's own needs. Are there any relevant guidelines?

The issue names no files, tests, or entry points. Start by identifying how deployment YAML expresses qwen3-14b, tp=2, replica=2, KV-cache offloading, and the requested vLLM version; done means a working example YAML and clear configuration guidelines.

Written by the indexing model from the issue text.