config.yaml

apiVersion: run.ai/v1
kind: RunaiJob
metadata:
  name: yyan # Name your pod. Start with your initials for example ep-pod-name. Must consist of lower case alphanumeric characters, '-' or '.', and must start and end with an alphanumeric character
  namespace: runai-ivrl-yyan # Use `runai list projects` to see what project you belong to. Then put the project namespace. It should be something like runai-ivrl-{your gaspar username}
  labels:
    user: yyan # Your username
spec:
  template:
    metadata:
      labels:
        user: yuhang.yan # User e.g. firstname.lastname
    spec:
      hostIPC: true
      schedulerName: runai-scheduler
      restartPolicy: Never
      nodeSelector:
        run.ai/type: G10 # "S8" (CPU only), "G9" (Nvidia V100) or "G10" (Nvidia A100)
      volumes:
        - name: data
          persistentVolumeClaim:
            claimName: runai-ivrl-yyan-ivrldata2 # Put your gaspar username here
        - name: ivrl-scratch
          persistentVolumeClaim:
            claimName: runai-ivrl-yyan-scratch # Put your gaspar username here
        - name: dshm
          emptyDir:
            medium: Memory
            sizeLimit: 4Gi # 4G shared memory allocated from memory.
      containers:
        - name: ubuntu # Name your container (pretty arbitrary)
          image: ic-registry.epfl.ch/ivrl/pajouheshgar/pytorch2.01:cuda11.7v2 # The docker image file you want to use. It should be uploaded to ic-registry
          env:
            - name: CLUSTER_USER
              value: yyan # Your epfl username. put inside ""
            - name: CLUSTER_USER_ID
              value: "286103" # Your epfl UID. put inside ""
            - name: CLUSTER_GROUP_NAME
              value: "ivrl" # Your group name. put inside ""
            - name: CLUSTER_GROUP_ID
              value: "30034" # Your epfl GID. put inside ""
          workingDir: /
          command: [ "/bin/bash", "-c" ] # Change ??? to your username bellow
          args: [ "source /opt/lab/setup.sh  && su yyan -c 'jupyter lab --ip=0.0.0.0 --no-browser --notebook-dir=/scratch'" ]
          ports:
            - containerPort: 8888
              name: jupyter
          imagePullPolicy: Always
          resources:
            requests:
              cpu: 16
              memory: "64Gi"
              nvidia.com/gpu: 2
            limits:
              nvidia.com/gpu: 2 # Number of GPU
          volumeMounts:
            - mountPath: /dev/shm
              name: dshm
            - mountPath: /scratch
              name: ivrl-scratch
            - mountPath: /data
              name: data