Interface Job

{Caffe2Settings} [caffe2Settings] Specifies the settings for Caffe2 job.

{CaffeSettings} [caffeSettings] Specifies the settings for Caffe job.

{ChainerSettings} [chainerSettings] Specifies the settings for Chainer job.

{ResourceId} [cluster] Cluster. Resource ID of the cluster associated with the job.

{CNTKsettings} [cntkSettings] Specifies the settings for CNTK (aka Microsoft Cognitive Toolkit) job.

{JobPropertiesConstraints} [constraints] Constraints associated with the Job.

{ContainerSettings} [containerSettings] If provided the job will run in the specified container. If the container was downloaded as part of cluster setup then the same container image will be used. If not provided, the job will run on the VM.

{Date} [creationTime] Creation time. The creation time of the job. NOTE: This property will not be serialized. It can only be populated by the server.

{CustomMpiSettings} [customMpiSettings] Specifies the settings for custom MPI job.

{CustomToolkitSettings} [customToolkitSettings] Specifies the settings for custom tool kit job.

{EnvironmentVariable[]} [environmentVariables] Environment variables. A collection of user defined environment variables to be setup for the job.

{JobPropertiesExecutionInfo} [executionInfo] Information about the execution of a job.

{ExecutionState} [executionState] Execution state. The current state of the job. Possible values are: queued - The job is queued and able to run. A job enters this state when it is created, or when it is awaiting a retry after a failed run. running - The job is running on a compute cluster. This includes job-level preparation such as downloading resource files or set up container specified on the job - it does not necessarily mean that the job command line has started executing. terminating - The job is terminated by the user, the terminate operation is in progress. succeeded - The job has completed running succesfully and exited with exit code 0. failed - The job has finished unsuccessfully (failed with a non-zero exit code) and has exhausted its retry limit. A job is also marked as failed if an error occurred launching the job. Possible values include: 'queued', 'running', 'terminating', 'succeeded', 'failed' NOTE: This property will not be serialized. It can only be populated by the server.

{Date} [executionStateTransitionTime] Execution state transition time. The time at which the job entered its current execution state. NOTE: This property will not be serialized. It can only be populated by the server.

{HorovodSettings} [horovodSettings] Specifies the settings for Horovod job.

{InputDirectory[]} [inputDirectories] Input directories. A list of input directories for the job.

{string} [jobOutputDirectoryPathSegment] Output directory path segment. A segment of job's output directories path created by Batch AI. Batch AI creates job's output directories under an unique path to avoid conflicts between jobs. This value contains a path segment generated by Batch AI to make the path unique and can be used to find the output directory on the node or mounted filesystem. NOTE: This property will not be serialized. It can only be populated by the server.

{JobPreparation} [jobPreparation] Specifies the actions to be performed before tool kit is launched. The specified actions will run on all the nodes that are part of the job

{MountVolumes} [mountVolumes] Mount volumes. Collection of mount volumes available to the job during execution. These volumes are mounted before the job execution and unmouted after the job completion. The volumes are mounted at location specified by $AZ_BATCHAI_JOB_MOUNT_ROOT environment variable.

{number} [nodeCount] Number of compute nodes to run the job on. The job will be gang scheduled on that many compute nodes

{OutputDirectory[]} [outputDirectories] Output directories. A list of output directories for the job.

{ProvisioningState} [provisioningState] Provisioning state. The provisioned state of the Batch AI job. Possible values include: 'creating', 'succeeded', 'failed', 'deleting' NOTE: This property will not be serialized. It can only be populated by the server.

{Date} [provisioningStateTransitionTime] Provisioning state transition time. The time at which the job entered its current provisioning state. NOTE: This property will not be serialized. It can only be populated by the server.

{PyTorchSettings} [pyTorchSettings] Specifies the settings for pyTorch job.

{JobPriority} [schedulingPriority] Scheduling priority. Scheduling priority associated with the job. Possible values include: 'low', 'normal', 'high'. Default value: 'normal' .

{EnvironmentVariableWithSecretValue[]} [secrets] Secrets. A collection of user defined environment variables with secret values to be setup for the job. Server will never report values of these variables back.

{string} [stdOutErrPathPrefix] Standard output directory path prefix. The path where the Batch AI service stores stdout, stderror and execution log of the job.

{TensorFlowSettings} [tensorFlowSettings] Specifies the settings for Tensor Flow job.

{ToolType} [toolType] The toolkit type of this job. Possible values are: cntk, tensorflow, caffe, caffe2, chainer, pytorch, custom, custommpi, horovod. Possible values include: 'cntk', 'tensorflow', 'caffe', 'caffe2', 'chainer', 'horovod', 'custommpi', 'custom'