CleanPodPolicyAll is the policy to clean all pods.

CleanPodPolicyNone is the policy to clean completed pods.

CleanPodPolicyRunning is the policy to clean all running pods.

CleanPodPolicyUndefined is no policy to clean pods.

FatalExitCode indicates a container exits with a unrecoverable error.

JobCreated means the job has been accepted by the system, but one or more of the pods/services has not been started.

JobFailed means one or more sub-resources (e.g.

JobPending means the job has been accepted by the system, but no pods/services has not been started.

JobRestarting means one or more sub-resources (e.g.

JobRunning means all sub-resources (e.g.

JobScaling means the job is scaling up/down its Pods.

JobSucceeded means all sub-resources (e.g.

ReasonOOMKilled indicates a container exits with OOM.

RestartPolicyAlways is to always restart pods.

RestartPolicyExitCode policy means that user should add exit code by themselves, The job operator will check these exit codes to determine the behavior when an error occurs: - 1-127: permanent error, do not restart.

RestartPolicyNever is to never restart pods.

RestartPolicyOnFailure is to restart pods on failure.

JobCondition describes the state of the job at a certain point.

JobStatus represents the current observed state of the training Job.

ReplicaSpec is a description of the replica.

ReplicaStatus represents the current observed state of the replica.

RunPolicy encapsulates various runtime policies of the distributed training job, for example how to clean up resources and how long the job can stay active.

SchedulingPolicy encapsulates various scheduling policies of the distributed training job, for example `minAvailable` for gang-scheduling.

CleanPodPolicy describes how to deal with pods when the job is finished.

JobConditionType defines all kinds of types of JobStatus.

ReplicaType represents the type of the replica.

RestartPolicy describes how the replicas should be restarted.