JobSettings

The ID of the user associated with the job.

The caller.

The pipeline ID.

Specifies whether to allow the job to use tidal resources. Valid values:

• true

• false

Specifies whether to enable fault tolerance monitoring for the job. Valid values:

• true

• false

The configuration parameters for fault tolerance monitoring after it is enabled, such as specifying whether to enable log hang-based detection.

Specifies whether to allow the job to use RDMA. Valid values:

• true

• false

Specifies whether to allow OSS append operations. Valid values:

• true

• false

The oversold resource usage mode for the job, which can be not accepted, accepted, or only accepted.

The additional advanced parameter settings.

The NVIDIA driver configuration.

Specifies whether to enable the computing power health check for the job. Valid values:

• true

• false

The configuration parameters for the computing power health check.

The retention duration after the job ends, in minutes.

The retention policy after the job ends.

The CPU affinity settings. This setting takes effect only when general-purpose computing subscription resources are used.

Specifies whether to skip the inventory check. Valid values:

• true

• false

Specifies whether to mount all RDMA NICs.

Specifies whether to enable the DSW development mode for the job. If enabled, a sidecar container is injected into the pod.

The DataJuicer job configuration.

The output model configuration. This parameter takes effect only in joint training scenarios.

Specifies whether to allow the job to be scheduled to nodes that are marked as unschedulable or included in the workspace blacklist. If enabled, node status checks are skipped and corresponding bypass configurations are applied.

The shell used to execute job commands, such as /bin/sh or /bin/bash.

The grace period to wait for the pod to exit when the job is terminated, in seconds.

The maximum number of restarts for an elastic spot job.

Specifies whether to allow job nodes to communicate with each other over SSH. The image must have SSHD built-in. This parameter takes effect only for PyTorch jobs that use self-managed computing resources. Valid values:

• true

• false