github.com/kubeflow/training-operator@v1.7.0/examples/mxnet/tune/mx_job_tune_gpu_v1.yaml (about)

     1  apiVersion: "kubeflow.org/v1"
     2  kind: "MXJob"
     3  metadata:
     4    name: "auto-tuning-job"
     5  spec:
     6    jobMode: MXTune
     7    mxReplicaSpecs:
     8      TunerTracker:
     9        replicas: 1
    10        restartPolicy: Never
    11        template:
    12          spec:
    13            containers:
    14            - name: mxnet
    15              image: mxjob/auto-tuning:gpu
    16              command: ["python3"]
    17              args: ["/home/scripts/start-job.py"]
    18      TunerServer:
    19        label: 2080ti
    20        replicas: 1
    21        restartPolicy: Never
    22        template:
    23          spec:
    24            containers:
    25            - name: mxnet
    26              image: mxjob/auto-tuning:gpu
    27              command: ["python3"]
    28              args: ["/home/scripts/start-job.py"]
    29              resources:
    30                limits:
    31                  nvidia.com/gpu: 1
    32      Tuner:
    33        replicas: 1
    34        restartPolicy: Never
    35        template:
    36          spec:
    37            containers:
    38            - name: mxnet
    39              image: mxjob/auto-tuning:gpu
    40              command: ["python3"]
    41              args: ["/home/scripts/start-job.py"]