sagemaker.core.workflow.pipeline_context

The pipeline context for workflow

Functions

retrieve_caller_name(job_instance)	Convenience method for runnable_by_pipeline decorator
runnable_by_pipeline(run_func)	A convenient Decorator

Classes

LocalPipelineSession([boto_session, ...])	Managing a session that executes Sagemaker pipelines and jobs locally in a pipeline context.
PipelineSession([boto_session, ...])	Managing interactions with SageMaker APIs and AWS services needed under Pipeline Context

class sagemaker.core.workflow.pipeline_context.LocalPipelineSession(boto_session=None, default_bucket=None, s3_endpoint_url=None, disable_local_code=False, default_bucket_prefix=None)

Bases: LocalSession, PipelineSession

Managing a session that executes Sagemaker pipelines and jobs locally in a pipeline context.

This class inherits from the LocalSession and PipelineSession classes. When running Sagemaker pipelines locally, this class is preferred over LocalSession.

class sagemaker.core.workflow.pipeline_context.PipelineSession(boto_session=None, sagemaker_client=None, default_bucket=None, settings=<sagemaker.core.session_settings.SessionSettings object>, sagemaker_config: dict | None = None, default_bucket_prefix: str | None = None)

Bases: Session

Managing interactions with SageMaker APIs and AWS services needed under Pipeline Context

This class inherits the SageMaker session, it provides convenient methods for manipulating entities and resources that Amazon SageMaker uses, such as training jobs, endpoints, and input datasets in S3. When composing SageMaker Model-Building Pipeline, PipelineSession is recommended over regular SageMaker Session

property context

Hold contextual information useful to the session

init_model_step_arguments(model)

Create a _ModelStepArguments (if not exist) as pipeline context

Parameters:

model (Model or PipelineModel) – A sagemaker.model.Model or sagemaker.pipeline.PipelineModel instance

sagemaker.core.workflow.pipeline_context.retrieve_caller_name(job_instance)

Convenience method for runnable_by_pipeline decorator

This function takes an instance of a job class and maps it to the pipeline session function that creates the job request.

Parameters:

job_instance – A job class instance, one of the following types: - Processor (from sagemaker.core.processing) - ModelTrainer (from sagemaker.train.model_trainer) - Transformer (from sagemaker.core.transformer) - HyperparameterTuner (from sagemaker.train.tuner)

Note

This function uses duck typing to avoid importing from Train package, which would create architecture violations (Core should not depend on Train). Instead of isinstance checks, we check for characteristic attributes/methods.

sagemaker.core.workflow.pipeline_context.runnable_by_pipeline(run_func)

A convenient Decorator

This is a decorator designed to annotate, during pipeline session, the methods that downstream managed to

1. preprocess user inputs, outputs, and configurations

2. generate the create request

3. start the job.

For instance, Processor.run, Estimator.fit, or Transformer.transform. This decorator will essentially run 1, and capture the request shape from 2, then instead of starting a new job in 3, it will return request shape from 2 to sagemaker.workflow.steps.Step. The request shape will be used to construct the arguments needed to compose that particular step as part of the pipeline. The job will be started during pipeline execution.