AWS Tools for Windows PowerShell
Command Reference

AWS services or capabilities described in AWS Documentation may vary by region/location. Click Getting Started with Amazon AWS to see specific differences applicable to the China (Beijing) Region.

Synopsis

Calls the SageMaker Training Session Runtime ExecuteTrainingOperation API operation.

Syntax

Invoke-SMTSRTrainingOperation
-OperationConfig_OptimizerStep_AdamParamsConfig_Beta1 <Double>
-OperationConfig_OptimizerStep_AdamParamsConfig_Beta2 <Double>
-OperationConfig_Forward_DatumList <Datum[]>
-OperationConfig_ForwardBackward_DatumList <Datum[]>
-OperationConfig_SaveState_Description <String>
-OperationConfig_SaveWeights_Description <String>
-OperationConfig_OptimizerStep_AdamParamsConfig_Ep <Double>
-OperationConfig_OptimizerStep_AdamParamsConfig_GradClipNorm <Double>
-OperationConfig_OptimizerStep_AdamParamsConfig_LearningRate <Double>
-OperationConfig_LoadState_LoadOptimizer <Boolean>
-OperationConfig_OptimizerStep_AdamParamsConfig_LoraPlusLrRatio <Double>
-OperationConfig_Forward_LossFnConfig <PSObject>
-OperationConfig_ForwardBackward_LossFnConfig <PSObject>
-OperationConfig_Forward_LossFnName <LossFnName>
-OperationConfig_ForwardBackward_LossFnName <LossFnName>
-OperationConfig_SaveWeights_MergeWeight <Boolean>
-OperationConfig_LoadState_ModelPackageArn <String>
-SequenceNumber <Int32>
-TrainingSessionName <String>
-OperationConfig_OptimizerStep_AdamParamsConfig_WeightDecay <Double>
-ClientToken <String>
-Select <String>
-Force <SwitchParameter>
-ClientConfig <AmazonSageMakerTrainingSessionRuntimeConfig>

Description

Submits an asynchronous training operation to the specified training session. The service executes operations in sequence number order. The session must be in Running or Blocked state. Use GetTrainingOperationResult to poll for the result. Supported operation types: forward, forwardBackward, optimizerStep, saveWeights, saveState, and loadState.

Parameters

-ClientConfig <AmazonSageMakerTrainingSessionRuntimeConfig>
Amazon.PowerShell.Cmdlets.SMTSR.AmazonSageMakerTrainingSessionRuntimeClientCmdlet.ClientConfig
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-ClientToken <String>
A unique, case-sensitive identifier that you provide to ensure that the operation completes no more than one time. If you retry a request with the same client token and the same parameters, the service ignores the request but does not return an error.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-Force <SwitchParameter>
This parameter overrides confirmation prompts to force the cmdlet to continue its operation. This parameter should always be used with caution.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_Forward_DatumList <Datum[]>
The list of input data to feed through the model. Must contain at least one datum, and each datum must have a non-empty modelInput with at least one chunk. Starting with version 4 of the SDK this property will default to null. If no data for this property is returned from the service the property will also be null. This was changed to improve performance and allow the SDK and caller to distinguish between a property not set or a property being empty to clear out a value. To retain the previous SDK behavior set the AWSConfigs.InitializeCollections static property to true.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_Forward_LossFnConfig <PSObject>
Additional loss function configuration passed to the engine. Supported keys are clip_low_threshold (valid range: 0 to 1) and clip_high_threshold (valid range: 1 to 10). Unsupported keys are rejected.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_Forward_LossFnName <LossFnName>
The name of the loss function to evaluate during the forward pass.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_ForwardBackward_DatumList <Datum[]>
The list of input data to feed through the model. Must contain at least one datum, and each datum must have a non-empty modelInput with at least one chunk. Starting with version 4 of the SDK this property will default to null. If no data for this property is returned from the service the property will also be null. This was changed to improve performance and allow the SDK and caller to distinguish between a property not set or a property being empty to clear out a value. To retain the previous SDK behavior set the AWSConfigs.InitializeCollections static property to true.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_ForwardBackward_LossFnConfig <PSObject>
Additional loss function configuration passed to the engine. Supported keys are clip_low_threshold (valid range: 0 to 1) and clip_high_threshold (valid range: 1 to 10). Unsupported keys are rejected.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_ForwardBackward_LossFnName <LossFnName>
The name of the loss function to evaluate during the forward-backward pass.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_LoadState_LoadOptimizer <Boolean>
Specifies whether to also restore the optimizer state alongside model weights. When true, optimizer state is restored to resume training from where it left off.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_LoadState_ModelPackageArn <String>
The ARN of the model package containing the checkpoint to load. Must be a valid SageMaker model package ARN in the format arn:aws:sagemaker:REGION:ACCOUNT:model-package/PACKAGE_NAME/VERSION. Model package group ARNs are not accepted.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_OptimizerStep_AdamParamsConfig_Beta1 <Double>
The exponential decay rate for the first moment estimates.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_OptimizerStep_AdamParamsConfig_Beta2 <Double>
The exponential decay rate for the second moment estimates.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_OptimizerStep_AdamParamsConfig_Ep <Double>
A small constant added to the denominator for numerical stability.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
AliasesOperationConfig_OptimizerStep_AdamParamsConfig_Eps
-OperationConfig_OptimizerStep_AdamParamsConfig_GradClipNorm <Double>
The maximum gradient norm for gradient clipping. Gradients are scaled down if their norm exceeds this value.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_OptimizerStep_AdamParamsConfig_LearningRate <Double>
The learning rate for the optimizer. Must be greater than 0.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_OptimizerStep_AdamParamsConfig_LoraPlusLrRatio <Double>
The learning rate multiplier for LoRA+ B matrices relative to A matrices. When set, B matrices use learningRate * loraPlusLrRatio while A matrices use the base learning rate.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_OptimizerStep_AdamParamsConfig_WeightDecay <Double>
The weight decay (L2 regularization) coefficient.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_SaveState_Description <String>
A human-readable description for the saved state checkpoint.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_SaveWeights_Description <String>
A human-readable description for the saved checkpoint.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-OperationConfig_SaveWeights_MergeWeight <Boolean>
Specifies whether to merge LoRA adapter weights into the base model before saving. When true, the saved checkpoint contains the full merged model weights.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
AliasesOperationConfig_SaveWeights_MergeWeights
-Select <String>
Use the -Select parameter to control the cmdlet output. The default value is 'OperationId'. Specifying -Select '*' will result in the cmdlet returning the whole service response (Amazon.SageMakerTrainingSessionRuntime.Model.ExecuteTrainingOperationResponse). Specifying the name of a property of type Amazon.SageMakerTrainingSessionRuntime.Model.ExecuteTrainingOperationResponse will result in that property being returned. Specifying -Select '^ParameterName' will result in the cmdlet returning the selected cmdlet parameter value.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-SequenceNumber <Int32>
The monotonically increasing sequence number that determines execution order. The training session processes operations strictly in sequence number order. If a gap exists, the session blocks until the missing operation arrives or times out. Must be at least 1.
Required?True
Position?Named
Accept pipeline input?True (ByPropertyName)
-TrainingSessionName <String>
The name of the training session to submit the operation to.
Required?True
Position?Named
Accept pipeline input?True (ByPropertyName)

Common Credential and Region Parameters

-AccessKey <String>
The AWS access key for the user account. This can be a temporary access key if the corresponding session token is supplied to the -SessionToken parameter.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
AliasesAK
-Credential <AWSCredentials>
An AWSCredentials object instance containing access and secret key information, and optionally a token for session-based credentials.
Required?False
Position?Named
Accept pipeline input?True (ByValue, ByPropertyName)
-EndpointUrl <String>
The endpoint to make the call against.Note: This parameter is primarily for internal AWS use and is not required/should not be specified for normal usage. The cmdlets normally determine which endpoint to call based on the region specified to the -Region parameter or set as default in the shell (via Set-DefaultAWSRegion). Only specify this parameter if you must direct the call to a specific custom endpoint.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
-NetworkCredential <PSCredential>
Used with SAML-based authentication when ProfileName references a SAML role profile. Contains the network credentials to be supplied during authentication with the configured identity provider's endpoint. This parameter is not required if the user's default network identity can or should be used during authentication.
Required?False
Position?Named
Accept pipeline input?True (ByValue, ByPropertyName)
-ProfileLocation <String>
Used to specify the name and location of the ini-format credential file (shared with the AWS CLI and other AWS SDKs)If this optional parameter is omitted this cmdlet will search the encrypted credential file used by the AWS SDK for .NET and AWS Toolkit for Visual Studio first. If the profile is not found then the cmdlet will search in the ini-format credential file at the default location: (user's home directory)\.aws\credentials.If this parameter is specified then this cmdlet will only search the ini-format credential file at the location given.As the current folder can vary in a shell or during script execution it is advised that you use specify a fully qualified path instead of a relative path.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
AliasesAWSProfilesLocation, ProfilesLocation
-ProfileName <String>
The user-defined name of an AWS credentials or SAML-based role profile containing credential information. The profile is expected to be found in the secure credential file shared with the AWS SDK for .NET and AWS Toolkit for Visual Studio. You can also specify the name of a profile stored in the .ini-format credential file used with the AWS CLI and other AWS SDKs.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
AliasesStoredCredentials, AWSProfileName
-Region <Object>
The system name of an AWS region or an AWSRegion instance. This governs the endpoint that will be used when calling service operations. Note that the AWS resources referenced in a call are usually region-specific.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
AliasesRegionToCall
-SecretKey <String>
The AWS secret key for the user account. This can be a temporary secret key if the corresponding session token is supplied to the -SessionToken parameter.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
AliasesSK, SecretAccessKey
-SessionToken <String>
The session token if the access and secret keys are temporary session-based credentials.
Required?False
Position?Named
Accept pipeline input?True (ByPropertyName)
AliasesST

Outputs

System.String or Amazon.SageMakerTrainingSessionRuntime.Model.ExecuteTrainingOperationResponse
This cmdlet returns a System.String object. The service call response (type Amazon.SageMakerTrainingSessionRuntime.Model.ExecuteTrainingOperationResponse) can be returned by specifying '-Select *'.
AWS Tools for PowerShell User Guide

Supported Version

AWS Tools for PowerShell: 2.x.y.z