name: 'PyTorch Operator - Training Job'
description: Kubeflow PyTorchJob launcher
inputs:
- {name: name,                       type: String,                          description: 'PyTorchJob name.'}
- {name: namespace,                  type: String,      default: kubeflow,  description: 'PyTorchJob namespace (likely your current namespace).'}
- {name: version,                    type: String,      default: v1,        description: 'PyTorchJob version.'}
- {name: master_spec,                type: JsonObject,  default: '{}',      description: 'PyTorchJob Master replicaSpecs.'}
- {name: worker_spec,                type: JsonObject,  default: '{}',      description: 'PyTorchJob Worker replicaSpecs.'}
- {name: job_timeout_minutes,        type: Integer,     default: 1440,      description: 'Time in minutes to wait for the job to complete.'}
- {name: delete_after_done,          type: Boolean,     default: 'True' ,   description: 'Whether to delete the job after it is finished.'}
- {name: clean_pod_policy,           type: String,      default: Running,   description: 'Defines the policy for cleaning up pods after the PyTorchJob completes.'}
- {name: active_deadline_seconds,    type: Integer,     optional: true,     description: 'Specifies the duration (in seconds) since startTime during which the job can remain active before it is terminated. Must be a positive integer. This setting applies only to pods where restartPolicy is OnFailure or Always.'}
- {name: backoff_limit,              type: Integer,     optional: true,     description: 'Number of retries before marking this job as failed.'}
- {name: ttl_seconds_after_finished, type: Integer,     optional: true,     description: 'Defines the TTL for cleaning up finished PyTorchJobs.'}
implementation:
  container:
    image: cascribner/kubeflow-pytorchjob-launcher:v1
    command: [python, /ml/launch_pytorchjob.py]
    args: 
      - --name
      - {inputValue: name}
      - --namespace
      - {inputValue: namespace}
      - --version
      - {inputValue: version}
      - --masterSpec
      - {inputValue: master_spec}
      - --workerSpec
      - {inputValue: worker_spec}
      - --jobTimeoutMinutes
      - {inputValue: job_timeout_minutes}
      - --deleteAfterDone
      - {inputValue: delete_after_done}
      - --cleanPodPolicy
      - {inputValue: clean_pod_policy}
      - if:
          cond: {isPresent: active_deadline_seconds}
          then:
          - --activeDeadlineSeconds
          - {inputValue: active_deadline_seconds}
      - if:
          cond: {isPresent: backoff_limit}
          then:
          - --backoffLimit
          - {inputValue: backoff_limit}
      - if:
          cond: {isPresent: ttl_seconds_after_finished}
          then:
          - --ttlSecondsAfterFinished
          - {inputValue: ttl_seconds_after_finished}