Skip to content

A library for configuring SageMaker pipelines using hierarchical configuration pattern.

License

Notifications You must be signed in to change notification settings

Nike-Inc/sagerender

Repository files navigation

SageRender


codecov Test PyPi Release Docker Build GitHub License Python Versions Docker Image Size Python Wheel

A library for configuring SageMaker pipelines using hierarchical configuration pattern. The purpose of this library is to isolate depdency on AWS SageMaker SDK and allow teams to create pipelines using configurations. The hierarchical configuration gives teams flexibility to use a lookups in configuration files to traverse through the hierarchy of configurations.

Table of Contents

Installation

From pypi:

pip install sagerender

From GitHub:

pip install git+https://github.com/nike-inc/sagerender#egg=sagerender

Changelog

See the changelog for a history of notable changes to SageRender.

Documentation

Commands

The most commonly used SageRender commands are:
run-pipeline		Run SageMaker Pipeline
upsert-pipeline		Configure and Upsert SageMaker Pipeline
version             Version of SageRender
Run Pipeline

Use this command to run an AWS SageMaker Pipeline

sagerender run-pipeline [-h] --pipeline-name PIPELINE_NAME
                               [--region-name REGION_NAME] [--verbose]

optional arguments:
  -h, --help            show this help message and exit
  --pipeline-name PIPELINE_NAME
                        Name of the SageMaker pipeline
  --region-name REGION_NAME
                        Name of the AWS Region
  --verbose, -v         Enable verbose logging
Upsert Pipeline

Use this command to create/update pipeline definition on AWS SageMaker.

usage: sagerender upsert-pipeline [-h] --pipeline-name PIPELINE_NAME [--max-parallel-execution-steps MAX_PARALLEL_EXECUTION_STEPS]
                                  [--default-bucket-prefix DEFAULT_BUCKET_PREFIX] [--dry-run] [--local] [--verbose]

options:
  -h, --help            show this help message and exit
  --pipeline-name PIPELINE_NAME
                        Name of the SageMaker pipeline
  --max-parallel-execution-steps MAX_PARALLEL_EXECUTION_STEPS
                        Set the max parallel execution steps for SageMaker pipeline
  --default-bucket-prefix DEFAULT_BUCKET_PREFIX
                        Default bucket prefix sets s3 path to be bucket-prefix/default-bucket-prefix/pipeline-name
  --dry-run             Skip upserting SageMaker pipeline definition. Used for debugging.
  --local, -l           Run in local pipeline mode
  --verbose, -v         Enable verbose logging

Getting Started

A notebook to help you get started with SageRender is available here.

Examples

  • A simple SageMaker Pipeline configuration with different environments can be found under examples/simple.
  • A complex SageMaker Pipeline configuration that defines different algorithms across different environments can be found under examples/complex.

Contribution

You can help by contributing to the project! Contribution guidelines can be found here. Instructions to easily set up your local development environment can be found here.

Unit Tests

Prerequisites

Setup

Install the package using poetry

poetry install

Run Tests

poetry run pytest

Run pre-commit checks

poetry run pre-commit run --all-files

Future Work

  • Add support for generating pipeline YAML templates
  • Add support for converting YAML to python code and vice versa.

About

A library for configuring SageMaker pipelines using hierarchical configuration pattern.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published