sagemaker

The idea is to rebuild AWS Sagemaker Python
SDK using
R6 classes and
paws behind the scenes.

Installation

You can install the development version of sagemaker from
GitHub with:

# install.packages("remotes)
remotes::install_github("DyfanJones/sagemaker-r-sdk")

Warning!!!

This repo is in constantly under development and is not currently
stable. sagemaker is currently aligning it’s api with sagemaker v2,
apologises for any code breaking this causes.

API overview:

This package aims to mimic python’s AWS Sagemaker SDK api, but using
R6 and paws

Architecture Design:

sagemaker is a metadata package that contains all methods to interact
with Amazon Sagemaker.

• sagemaker.core:
  Containse core components of sdk for example Session R6 class
• sagemaker.common:
  Contains common components used throughout sagemaker sdk
• sagemaker.mlcore:
  Contains core components for machine learning (ML) and amazon
  developed ML.
• sagemaker.mlframework:
  Contains ML frameworks developed for Amazon Sagemaker i.e. SKLearn
• sagemaker.workflow:
  Contains sagemaker pipeline and workflows
• sagemaker.debugger:
  Contains debugging methods
  (https://github.com/awslabs/sagemaker-debugger-rulesconfig)

Learn from examples:

Amazon Algorithms:

sagemaker is designed to minic python’s sagemaker sdk. Therefore all
examples for python’s sagemaker should be able to accessible.

Examples:

• Targeted Direct
  Marketing
  predicts potential customers that are most likely to convert based
  on customer and aggregate level metrics, using Amazon SageMaker’s
  implementation of XGBoost.
• XGBoost
  Tuning
  shows how to use SageMaker hyperparameter tuning to improve your
  model fits for the Targeted Direct
  Marketing
  task.
• BlazingText
  Word2Vec
  generates Word2Vec embeddings from a cleaned text dump of Wikipedia
  articles using SageMaker’s fast and scalable BlazingText
  implementation.

R Model Examples:

• R Multivariate Adaptive Regression
  Splines
  example over iris data.frame

Note: If a feature hasn’t yet been implemented please feel free to
raise a pull request or a ticket

For developers

To keep the package within the CRAN size limit of 5MB. sagemaker is
currently using a separate repository
(sagemaker-r-test-data)
to store R variants of test data stored in
sagemaker-python-sdk.
sagemaker-r-test-data will only consist of data that can’t be read into
R natively i.e. python pickle files. For other test data sagemaker will
read it directly from sagemaker-python-sdk.