- Can we deploy and run Qwen models inside our own Kubernetes cluster (self-hosted environment), and if yes:
- What are the recommended deployment methods? (e.g., Helm charts, Docker containers, vLLM, Ollama, HuggingFace TGI, Model Serving frameworks, etc.)
- Any prerequisites regarding GPU requirements or inference performance tips?
If anyone has experience deploying Qwen models on Kubernetes, could you share best practices or references?