apiVersion: lab.modeldock.dev/v1alpha1
kind: ModelService
metadata: {name: qwen-runtime, namespace: modeldock-lab}
spec:
  image: vllm/vllm-openai:v0.10.2
  runtime: vllm
  model: Qwen/Qwen2.5-1.5B-Instruct
  gpuPerReplica: 1
  replicas: 1
  cachePVC: model-weights