tutorials-and-examples/nvidia-bionemo/pretraining/job.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: bionemo-training
  namespace: bionemo-training
spec:
  template:
    spec:
      containers:
      - name: training
        image: nvcr.io/nvidia/clara/bionemo-framework:2.1
        command: ["/bin/bash"]
        args: ["/scripts/start.sh"]
        resources:
          limits:
            nvidia.com/gpu: 1
        volumeMounts:
        - name: bionemo-storage
          mountPath: /mnt/data
          subPath: training-data
        - name: bionemo-storage
          mountPath: /workspace/bionemo2/results
          subPath: tensorboard-logs
        - name: config-volume
          mountPath: /config
        - name: script-volume
          mountPath: /scripts
      volumes:
      - name: bionemo-storage
        persistentVolumeClaim:
          claimName: bionemo-filestore
      - name: config-volume
        configMap:
          name: bionemo-training-config
      - name: script-volume
        configMap:
          name: bionemo-startup-script
          defaultMode: 0777
      restartPolicy: Never
      nodeSelector:
        cloud.google.com/gke-gpu: "true"