Skip to content

ProcessingInstancePreference

Source code in docs_src/capo_sagemaker/types/processing_instance_preference.py 
12
13
14
15
16
17
18
19
20
class ProcessingInstancePreference(TypedDict, closed=True):
    instance_type: NotRequired[
        "capo_sagemaker.types.processing_instance_type.ProcessingInstanceType"
    ]
    """<p>The ML compute instance type. An instance type can appear only once in an <code>InstancePreferences</code> list.</p>"""
    instance_count: NotRequired[
        "capo_sagemaker.types.processing_instance_count.ProcessingInstanceCount"
    ]
    """<p>The number of instances to launch if this instance type is selected. Specify the instance count for the processing job in one of the following two ways:</p> <ol> <li> <p> <b>Per preference</b> – Set <code>InstanceCount</code> on every preference in the <code>InstancePreferences</code> list and don't set <code>ProcessingClusterConfig$InstanceCount</code>. Use this when each instance type needs a different number of instances to deliver equivalent compute.</p> </li> <li> <p> <b>One count for the job</b> – Set <code>ProcessingClusterConfig$InstanceCount</code> and omit it from every preference. Amazon SageMaker applies this to all instance types in the list.</p> </li> </ol> <p>For example, in a list of five preferences, either all five specify <code>InstanceCount</code> or none of them do. Amazon SageMaker rejects requests that set <code>InstanceCount</code> on only some preferences, that set it both per preference and in <code>ProcessingClusterConfig</code>, or that omit it in both places.</p>"""

instance_count instance-attribute 

instance_count: NotRequired[ProcessingInstanceCount]

The number of instances to launch if this instance type is selected. Specify the instance count for the processing job in one of the following two ways:

  1. Per preference – Set InstanceCount on every preference in the InstancePreferences list and don't set ProcessingClusterConfig$InstanceCount. Use this when each instance type needs a different number of instances to deliver equivalent compute.

  2. One count for the job – Set ProcessingClusterConfig$InstanceCount and omit it from every preference. Amazon SageMaker applies this to all instance types in the list.

For example, in a list of five preferences, either all five specify InstanceCount or none of them do. Amazon SageMaker rejects requests that set InstanceCount on only some preferences, that set it both per preference and in ProcessingClusterConfig, or that omit it in both places.

instance_type instance-attribute 

instance_type: NotRequired[ProcessingInstanceType]

The ML compute instance type. An instance type can appear only once in an InstancePreferences list.