apiVersion: v1
kind: PersistentVolumeClaim
metadata: {name: model-weights, namespace: modeldock-lab}
spec:
  accessModes: [ReadWriteOnce]
  resources: {requests: {storage: 20Gi}}
---
apiVersion: apps/v1
kind: Deployment
metadata: {name: qwen-gpu, namespace: modeldock-lab}
spec:
  strategy: {type: Recreate}
  replicas: 1
  selector: {matchLabels: {app: qwen-gpu}}
  template:
    metadata: {labels: {app: qwen-gpu}}
    spec:
      containers:
      - name: inference
        image: vllm/vllm-openai:v0.10.2
        args: [--model, Qwen/Qwen2.5-1.5B-Instruct, --served-model-name, qwen-gpu, --max-model-len, "4096"]
        ports: [{containerPort: 8000}]
        resources: {requests: {cpu: "2", memory: 4Gi, nvidia.com/gpu: 1}, limits: {memory: 12Gi, nvidia.com/gpu: 1}}
        startupProbe: {httpGet: {path: /health, port: 8000}, periodSeconds: 10, failureThreshold: 90}
        readinessProbe: {httpGet: {path: /health, port: 8000}, periodSeconds: 5}
        volumeMounts: [{name: weights, mountPath: /root/.cache/huggingface}, {name: shm, mountPath: /dev/shm}]
      volumes: [{name: weights, persistentVolumeClaim: {claimName: model-weights}}, {name: shm, emptyDir: {medium: Memory, sizeLimit: 1Gi}}]
---
apiVersion: v1
kind: Service
metadata: {name: qwen-gpu, namespace: modeldock-lab}
spec: {selector: {app: qwen-gpu}, ports: [{name: http, port: 8000, targetPort: 8000}]}