GPU4.yaml

#Below sample yaml is used to schedule 4 GPU pods
apiVersion: v1
kind: Pod
metadata:
 name: sfr-pod-yuewang-vdbert
 namespace: sfr-ns-yuewang
spec:
  restartPolicy: OnFailure
  volumes:
    - name: sfr-home-pv-yuewang
      persistentVolumeClaim:
        claimName: sfr-home-pvc-yuewang
    - name: sfr-share-pv-yuewang
      persistentVolumeClaim:
        claimName: sfr-share-pvc-yuewang
    - name: dshm
      emptyDir:
        medium: Memory
  containers:
    - name: vlp4
      image: gcr.io/salesforce-research-internal/yuewang-vlp-new
      command: ["sleep", "infinity"]
      resources:
        limits:
          nvidia.com/gpu: 4
          cpu: "31"
          memory: 192G
      volumeMounts:
        - name: sfr-home-pv-yuewang
          mountPath: "/export/home"
        - name: sfr-share-pv-yuewang
          mountPath: "/export/share"
        - mountPath: "/dev/shm"
          name: dshm

  tolerations:
  - key: "gpu_num"
    operator: "Equal"
    value: "8"
    effect: "NoSchedule"