Sdílet prostřednictvím


ArmMachineLearningModelFactory.SparkJob Method

Definition

Overloads

SparkJob(String, IDictionary<String,String>, IDictionary<String, String>, ResourceIdentifier, ResourceIdentifier, String, String, MachineLearningIdentityConfiguration, Nullable<Boolean>, NotificationSetting, IDictionary<String,SecretConfiguration>, IDictionary<String, MachineLearningJobService>, Nullable<MachineLearningJobStatus>, IEnumerable<String>, String, String, IDictionary<String,String>, SparkJobEntry, String, IEnumerable<String>, IDictionary<String, MachineLearningJobInput>, IEnumerable<String>, IDictionary<String, MachineLearningJobOutput>, IEnumerable<String>, JobQueueSettings, SparkResourceConfiguration)

Initializes a new instance of SparkJob.

SparkJob(String, IDictionary<String,String>, IDictionary<String, String>, String, Nullable<MachineLearningJobStatus>, String, IDictionary<String, MachineLearningJobService>, ResourceIdentifier, Nullable<Boolean>, MachineLearningIdentityConfiguration, ResourceIdentifier, NotificationSetting, SparkResourceConfiguration, String, ResourceIdentifier, SparkJobEntry, ResourceIdentifier, IDictionary<String,MachineLearningJobInput>, IDictionary<String,MachineLearningJobOutput>, IEnumerable<String>, IEnumerable<String>, IEnumerable<String>, IEnumerable<String>, IDictionary<String,String>, Nullable<JobTier>, IDictionary<String, String>)

Initializes a new instance of SparkJob.

SparkJob(String, IDictionary<String,String>, IDictionary<String, String>, ResourceIdentifier, ResourceIdentifier, String, String, MachineLearningIdentityConfiguration, Nullable<Boolean>, NotificationSetting, IDictionary<String,SecretConfiguration>, IDictionary<String, MachineLearningJobService>, Nullable<MachineLearningJobStatus>, IEnumerable<String>, String, String, IDictionary<String,String>, SparkJobEntry, String, IEnumerable<String>, IDictionary<String, MachineLearningJobInput>, IEnumerable<String>, IDictionary<String, MachineLearningJobOutput>, IEnumerable<String>, JobQueueSettings, SparkResourceConfiguration)

Source:
ArmMachineLearningModelFactory.cs

Initializes a new instance of SparkJob.

public static Azure.ResourceManager.MachineLearning.Models.SparkJob SparkJob (string description = default, System.Collections.Generic.IDictionary<string,string> properties = default, System.Collections.Generic.IDictionary<string,string> tags = default, Azure.Core.ResourceIdentifier componentId = default, Azure.Core.ResourceIdentifier computeId = default, string displayName = default, string experimentName = default, Azure.ResourceManager.MachineLearning.Models.MachineLearningIdentityConfiguration identity = default, bool? isArchived = default, Azure.ResourceManager.MachineLearning.Models.NotificationSetting notificationSetting = default, System.Collections.Generic.IDictionary<string,Azure.ResourceManager.MachineLearning.Models.SecretConfiguration> secretsConfiguration = default, System.Collections.Generic.IDictionary<string,Azure.ResourceManager.MachineLearning.Models.MachineLearningJobService> services = default, Azure.ResourceManager.MachineLearning.Models.MachineLearningJobStatus? status = default, System.Collections.Generic.IEnumerable<string> archives = default, string args = default, string codeId = default, System.Collections.Generic.IDictionary<string,string> conf = default, Azure.ResourceManager.MachineLearning.Models.SparkJobEntry entry = default, string environmentId = default, System.Collections.Generic.IEnumerable<string> files = default, System.Collections.Generic.IDictionary<string,Azure.ResourceManager.MachineLearning.Models.MachineLearningJobInput> inputs = default, System.Collections.Generic.IEnumerable<string> jars = default, System.Collections.Generic.IDictionary<string,Azure.ResourceManager.MachineLearning.Models.MachineLearningJobOutput> outputs = default, System.Collections.Generic.IEnumerable<string> pyFiles = default, Azure.ResourceManager.MachineLearning.Models.JobQueueSettings queueSettings = default, Azure.ResourceManager.MachineLearning.Models.SparkResourceConfiguration resources = default);
static member SparkJob : string * System.Collections.Generic.IDictionary<string, string> * System.Collections.Generic.IDictionary<string, string> * Azure.Core.ResourceIdentifier * Azure.Core.ResourceIdentifier * string * string * Azure.ResourceManager.MachineLearning.Models.MachineLearningIdentityConfiguration * Nullable<bool> * Azure.ResourceManager.MachineLearning.Models.NotificationSetting * System.Collections.Generic.IDictionary<string, Azure.ResourceManager.MachineLearning.Models.SecretConfiguration> * System.Collections.Generic.IDictionary<string, Azure.ResourceManager.MachineLearning.Models.MachineLearningJobService> * Nullable<Azure.ResourceManager.MachineLearning.Models.MachineLearningJobStatus> * seq<string> * string * string * System.Collections.Generic.IDictionary<string, string> * Azure.ResourceManager.MachineLearning.Models.SparkJobEntry * string * seq<string> * System.Collections.Generic.IDictionary<string, Azure.ResourceManager.MachineLearning.Models.MachineLearningJobInput> * seq<string> * System.Collections.Generic.IDictionary<string, Azure.ResourceManager.MachineLearning.Models.MachineLearningJobOutput> * seq<string> * Azure.ResourceManager.MachineLearning.Models.JobQueueSettings * Azure.ResourceManager.MachineLearning.Models.SparkResourceConfiguration -> Azure.ResourceManager.MachineLearning.Models.SparkJob
Public Shared Function SparkJob (Optional description As String = Nothing, Optional properties As IDictionary(Of String, String) = Nothing, Optional tags As IDictionary(Of String, String) = Nothing, Optional componentId As ResourceIdentifier = Nothing, Optional computeId As ResourceIdentifier = Nothing, Optional displayName As String = Nothing, Optional experimentName As String = Nothing, Optional identity As MachineLearningIdentityConfiguration = Nothing, Optional isArchived As Nullable(Of Boolean) = Nothing, Optional notificationSetting As NotificationSetting = Nothing, Optional secretsConfiguration As IDictionary(Of String, SecretConfiguration) = Nothing, Optional services As IDictionary(Of String, MachineLearningJobService) = Nothing, Optional status As Nullable(Of MachineLearningJobStatus) = Nothing, Optional archives As IEnumerable(Of String) = Nothing, Optional args As String = Nothing, Optional codeId As String = Nothing, Optional conf As IDictionary(Of String, String) = Nothing, Optional entry As SparkJobEntry = Nothing, Optional environmentId As String = Nothing, Optional files As IEnumerable(Of String) = Nothing, Optional inputs As IDictionary(Of String, MachineLearningJobInput) = Nothing, Optional jars As IEnumerable(Of String) = Nothing, Optional outputs As IDictionary(Of String, MachineLearningJobOutput) = Nothing, Optional pyFiles As IEnumerable(Of String) = Nothing, Optional queueSettings As JobQueueSettings = Nothing, Optional resources As SparkResourceConfiguration = Nothing) As SparkJob

Parameters

description
String

The asset description text.

properties
IDictionary<String,String>

The asset property dictionary.

tags
IDictionary<String,String>

Tag dictionary. Tags can be added, removed, and updated.

componentId
ResourceIdentifier

ARM resource ID of the component resource.

computeId
ResourceIdentifier

ARM resource ID of the compute resource.

displayName
String

Display name of job.

experimentName
String

The name of the experiment the job belongs to. If not set, the job is placed in the "Default" experiment.

identity
MachineLearningIdentityConfiguration

Identity configuration. If set, this should be one of AmlToken, ManagedIdentity, UserIdentity or null. Defaults to AmlToken if null. Please note MachineLearningIdentityConfiguration is the base class. According to the scenario, a derived class of the base class might need to be assigned here, or this property needs to be casted to one of the possible derived classes. The available derived classes include AmlToken, MachineLearningManagedIdentity and MachineLearningUserIdentity.

isArchived
Nullable<Boolean>

Is the asset archived?.

notificationSetting
NotificationSetting

Notification setting for the job.

secretsConfiguration
IDictionary<String,SecretConfiguration>

Configuration for secrets to be made available during runtime.

services
IDictionary<String,MachineLearningJobService>

List of JobEndpoints. For local jobs, a job endpoint will have an endpoint value of FileStreamObject.

status
Nullable<MachineLearningJobStatus>

Status of the job.

archives
IEnumerable<String>

Archive files used in the job.

args
String

Arguments for the job.

codeId
String

[Required] ARM resource ID of the code asset.

conf
IDictionary<String,String>

Spark configured properties.

entry
SparkJobEntry

[Required] The entry to execute on startup of the job. Please note SparkJobEntry is the base class. According to the scenario, a derived class of the base class might need to be assigned here, or this property needs to be casted to one of the possible derived classes. The available derived classes include SparkJobPythonEntry and SparkJobScalaEntry.

environmentId
String

The ARM resource ID of the Environment specification for the job.

files
IEnumerable<String>

Files used in the job.

inputs
IDictionary<String,MachineLearningJobInput>

Mapping of input data bindings used in the job. Please note MachineLearningJobInput is the base class. According to the scenario, a derived class of the base class might need to be assigned here, or this property needs to be casted to one of the possible derived classes. The available derived classes include MachineLearningCustomModelJobInput, MachineLearningLiteralJobInput, MachineLearningFlowModelJobInput, MachineLearningTableJobInput, MachineLearningTritonModelJobInput, MachineLearningUriFileJobInput and MachineLearningUriFolderJobInput.

jars
IEnumerable<String>

Jar files used in the job.

outputs
IDictionary<String,MachineLearningJobOutput>

Mapping of output data bindings used in the job. Please note MachineLearningJobOutput is the base class. According to the scenario, a derived class of the base class might need to be assigned here, or this property needs to be casted to one of the possible derived classes. The available derived classes include MachineLearningCustomModelJobOutput, MachineLearningFlowModelJobOutput, MachineLearningTableJobOutput, MachineLearningTritonModelJobOutput, MachineLearningUriFileJobOutput and MachineLearningUriFolderJobOutput.

pyFiles
IEnumerable<String>

Python files used in the job.

queueSettings
JobQueueSettings

Queue settings for the job.

resources
SparkResourceConfiguration

Compute Resource configuration for the job.

Returns

A new SparkJob instance for mocking.

Applies to

SparkJob(String, IDictionary<String,String>, IDictionary<String, String>, String, Nullable<MachineLearningJobStatus>, String, IDictionary<String, MachineLearningJobService>, ResourceIdentifier, Nullable<Boolean>, MachineLearningIdentityConfiguration, ResourceIdentifier, NotificationSetting, SparkResourceConfiguration, String, ResourceIdentifier, SparkJobEntry, ResourceIdentifier, IDictionary<String,MachineLearningJobInput>, IDictionary<String,MachineLearningJobOutput>, IEnumerable<String>, IEnumerable<String>, IEnumerable<String>, IEnumerable<String>, IDictionary<String,String>, Nullable<JobTier>, IDictionary<String, String>)

Source:
ArmMachineLearningModelFactory.cs

Initializes a new instance of SparkJob.

public static Azure.ResourceManager.MachineLearning.Models.SparkJob SparkJob (string description = default, System.Collections.Generic.IDictionary<string,string> tags = default, System.Collections.Generic.IDictionary<string,string> properties = default, string displayName = default, Azure.ResourceManager.MachineLearning.Models.MachineLearningJobStatus? status = default, string experimentName = default, System.Collections.Generic.IDictionary<string,Azure.ResourceManager.MachineLearning.Models.MachineLearningJobService> services = default, Azure.Core.ResourceIdentifier computeId = default, bool? isArchived = default, Azure.ResourceManager.MachineLearning.Models.MachineLearningIdentityConfiguration identity = default, Azure.Core.ResourceIdentifier componentId = default, Azure.ResourceManager.MachineLearning.Models.NotificationSetting notificationSetting = default, Azure.ResourceManager.MachineLearning.Models.SparkResourceConfiguration resources = default, string args = default, Azure.Core.ResourceIdentifier codeId = default, Azure.ResourceManager.MachineLearning.Models.SparkJobEntry entry = default, Azure.Core.ResourceIdentifier environmentId = default, System.Collections.Generic.IDictionary<string,Azure.ResourceManager.MachineLearning.Models.MachineLearningJobInput> inputs = default, System.Collections.Generic.IDictionary<string,Azure.ResourceManager.MachineLearning.Models.MachineLearningJobOutput> outputs = default, System.Collections.Generic.IEnumerable<string> pyFiles = default, System.Collections.Generic.IEnumerable<string> jars = default, System.Collections.Generic.IEnumerable<string> files = default, System.Collections.Generic.IEnumerable<string> archives = default, System.Collections.Generic.IDictionary<string,string> conf = default, Azure.ResourceManager.MachineLearning.Models.JobTier? queueJobTier = default, System.Collections.Generic.IDictionary<string,string> environmentVariables = default);
static member SparkJob : string * System.Collections.Generic.IDictionary<string, string> * System.Collections.Generic.IDictionary<string, string> * string * Nullable<Azure.ResourceManager.MachineLearning.Models.MachineLearningJobStatus> * string * System.Collections.Generic.IDictionary<string, Azure.ResourceManager.MachineLearning.Models.MachineLearningJobService> * Azure.Core.ResourceIdentifier * Nullable<bool> * Azure.ResourceManager.MachineLearning.Models.MachineLearningIdentityConfiguration * Azure.Core.ResourceIdentifier * Azure.ResourceManager.MachineLearning.Models.NotificationSetting * Azure.ResourceManager.MachineLearning.Models.SparkResourceConfiguration * string * Azure.Core.ResourceIdentifier * Azure.ResourceManager.MachineLearning.Models.SparkJobEntry * Azure.Core.ResourceIdentifier * System.Collections.Generic.IDictionary<string, Azure.ResourceManager.MachineLearning.Models.MachineLearningJobInput> * System.Collections.Generic.IDictionary<string, Azure.ResourceManager.MachineLearning.Models.MachineLearningJobOutput> * seq<string> * seq<string> * seq<string> * seq<string> * System.Collections.Generic.IDictionary<string, string> * Nullable<Azure.ResourceManager.MachineLearning.Models.JobTier> * System.Collections.Generic.IDictionary<string, string> -> Azure.ResourceManager.MachineLearning.Models.SparkJob
Public Shared Function SparkJob (Optional description As String = Nothing, Optional tags As IDictionary(Of String, String) = Nothing, Optional properties As IDictionary(Of String, String) = Nothing, Optional displayName As String = Nothing, Optional status As Nullable(Of MachineLearningJobStatus) = Nothing, Optional experimentName As String = Nothing, Optional services As IDictionary(Of String, MachineLearningJobService) = Nothing, Optional computeId As ResourceIdentifier = Nothing, Optional isArchived As Nullable(Of Boolean) = Nothing, Optional identity As MachineLearningIdentityConfiguration = Nothing, Optional componentId As ResourceIdentifier = Nothing, Optional notificationSetting As NotificationSetting = Nothing, Optional resources As SparkResourceConfiguration = Nothing, Optional args As String = Nothing, Optional codeId As ResourceIdentifier = Nothing, Optional entry As SparkJobEntry = Nothing, Optional environmentId As ResourceIdentifier = Nothing, Optional inputs As IDictionary(Of String, MachineLearningJobInput) = Nothing, Optional outputs As IDictionary(Of String, MachineLearningJobOutput) = Nothing, Optional pyFiles As IEnumerable(Of String) = Nothing, Optional jars As IEnumerable(Of String) = Nothing, Optional files As IEnumerable(Of String) = Nothing, Optional archives As IEnumerable(Of String) = Nothing, Optional conf As IDictionary(Of String, String) = Nothing, Optional queueJobTier As Nullable(Of JobTier) = Nothing, Optional environmentVariables As IDictionary(Of String, String) = Nothing) As SparkJob

Parameters

description
String

The asset description text.

tags
IDictionary<String,String>

Tag dictionary. Tags can be added, removed, and updated.

properties
IDictionary<String,String>

The asset property dictionary.

displayName
String

Display name of job.

status
Nullable<MachineLearningJobStatus>

Status of the job.

experimentName
String

The name of the experiment the job belongs to. If not set, the job is placed in the "Default" experiment.

services
IDictionary<String,MachineLearningJobService>

List of JobEndpoints. For local jobs, a job endpoint will have an endpoint value of FileStreamObject.

computeId
ResourceIdentifier

ARM resource ID of the compute resource.

isArchived
Nullable<Boolean>

Is the asset archived?.

identity
MachineLearningIdentityConfiguration

Identity configuration. If set, this should be one of AmlToken, ManagedIdentity, UserIdentity or null. Defaults to AmlToken if null. Please note MachineLearningIdentityConfiguration is the base class. According to the scenario, a derived class of the base class might need to be assigned here, or this property needs to be casted to one of the possible derived classes. The available derived classes include AmlToken, MachineLearningManagedIdentity and MachineLearningUserIdentity.

componentId
ResourceIdentifier

ARM resource ID of the component resource.

notificationSetting
NotificationSetting

Notification setting for the job.

resources
SparkResourceConfiguration

Compute Resource configuration for the job.

args
String

Arguments for the job.

codeId
ResourceIdentifier

[Required] arm-id of the code asset.

entry
SparkJobEntry

[Required] The entry to execute on startup of the job. Please note SparkJobEntry is the base class. According to the scenario, a derived class of the base class might need to be assigned here, or this property needs to be casted to one of the possible derived classes. The available derived classes include SparkJobPythonEntry and SparkJobScalaEntry.

environmentId
ResourceIdentifier

The ARM resource ID of the Environment specification for the job.

inputs
IDictionary<String,MachineLearningJobInput>

Mapping of input data bindings used in the job. Please note MachineLearningJobInput is the base class. According to the scenario, a derived class of the base class might need to be assigned here, or this property needs to be casted to one of the possible derived classes. The available derived classes include MachineLearningCustomModelJobInput, MachineLearningLiteralJobInput, MachineLearningFlowModelJobInput, MachineLearningTableJobInput, MachineLearningTritonModelJobInput, MachineLearningUriFileJobInput and MachineLearningUriFolderJobInput.

outputs
IDictionary<String,MachineLearningJobOutput>

Mapping of output data bindings used in the job. Please note MachineLearningJobOutput is the base class. According to the scenario, a derived class of the base class might need to be assigned here, or this property needs to be casted to one of the possible derived classes. The available derived classes include MachineLearningCustomModelJobOutput, MachineLearningFlowModelJobOutput, MachineLearningTableJobOutput, MachineLearningTritonModelJobOutput, MachineLearningUriFileJobOutput and MachineLearningUriFolderJobOutput.

pyFiles
IEnumerable<String>

Python files used in the job.

jars
IEnumerable<String>

Jar files used in the job.

files
IEnumerable<String>

Files used in the job.

archives
IEnumerable<String>

Archive files used in the job.

conf
IDictionary<String,String>

Spark configured properties.

queueJobTier
Nullable<JobTier>

Queue settings for the job.

environmentVariables
IDictionary<String,String>

Environment variables included in the job.

Returns

A new SparkJob instance for mocking.

Applies to