SageMakerMetrics

Table of Contents

• SageMakerMetrics
  • Client
  • Paginators

Client

class SageMakerMetrics.Client

A low-level client representing Amazon SageMaker Metrics Service

Contains all data plane API operations and data types for Amazon SageMaker Metrics. Use these APIs to put and retrieve (get) features related to your training run.

• BatchPutMetrics

import boto3

client = boto3.client('sagemaker-metrics')

These are the available methods:

• batch_put_metrics()
• can_paginate()
• close()
• get_paginator()
• get_waiter()

batch_put_metrics(**kwargs)

Used to ingest training metrics into SageMaker. These metrics can be visualized in SageMaker Studio and retrieved with the GetMetrics API.

See also: AWS API Documentation

Request Syntax

response = client.batch_put_metrics(
    TrialComponentName='string',
    MetricData=[
        {
            'MetricName': 'string',
            'Timestamp': datetime(2015, 1, 1),
            'Step': 123,
            'Value': 123.0
        },
    ]
)

Parameters

• TrialComponentName (string) --

  [REQUIRED]

  The name of the Trial Component to associate with the metrics.

• MetricData (list) --

  [REQUIRED]

  A list of raw metric values to put.

  • (dict) --

    The raw metric data to associate with the resource.

    • MetricName (string) -- [REQUIRED]

      The name of the metric.

    • Timestamp (datetime) -- [REQUIRED]

      The time that the metric was recorded.

    • Step (integer) --

      The metric step (epoch).

    • Value (float) -- [REQUIRED]

      The metric value.

Return type

dict

Returns

Response Syntax

{
    'Errors': [
        {
            'Code': 'METRIC_LIMIT_EXCEEDED'|'INTERNAL_ERROR'|'VALIDATION_ERROR'|'CONFLICT_ERROR',
            'MetricIndex': 123
        },
    ]
}

Response Structure

• (dict) --

  • Errors (list) --

    Lists any errors that occur when inserting metric data.

    • (dict) --

      An error that occured when putting the metric data.

      • Code (string) --

        The error code of an error that occured when attempting to put metrics.

        • METRIC_LIMIT_EXCEEDED : The maximum amount of metrics per resource is exceeded.
        • INTERNAL_ERROR : An internal error occured.
        • VALIDATION_ERROR : The metric data failed validation.
        • CONFLICT_ERROR : Multiple requests attempted to modify the same data simultaneously.

      • MetricIndex (integer) --

        An index that corresponds to the metric in the request.

can_paginate(operation_name)

Check if an operation can be paginated.

Parameters

operation_name (string) -- The operation name. This is the same name as the method name on the client. For example, if the method name is create_foo, and you'd normally invoke the operation as client.create_foo(**kwargs), if the create_foo operation can be paginated, you can use the call client.get_paginator("create_foo").

Returns

True if the operation can be paginated, False otherwise.

close()

Closes underlying endpoint connections.

get_paginator(operation_name)

Create a paginator for an operation.

Parameters

operation_name (string) -- The operation name. This is the same name as the method name on the client. For example, if the method name is create_foo, and you'd normally invoke the operation as client.create_foo(**kwargs), if the create_foo operation can be paginated, you can use the call client.get_paginator("create_foo").

Raises OperationNotPageableError

Raised if the operation is not pageable. You can use the client.can_paginate method to check if an operation is pageable.

Return type

L{botocore.paginate.Paginator}

Returns

A paginator object.

get_waiter(waiter_name)

Returns an object that can wait for some condition.

Parameters

waiter_name (str) -- The name of the waiter to get. See the waiters section of the service docs for a list of available waiters.

Returns

The specified waiter object.

Return type

botocore.waiter.Waiter

Paginators

The available paginators are: