Skip to content

Example of how to deploy an ML algorithm together with SHAP explanations to AWS Sagemaker, including a front end dashboard.

Notifications You must be signed in to change notification settings

oegedijk/sagemaker-creditscore-explainer

Repository files navigation

README

by: Oege Dijk

This repository is a demonstration of how to deploy explainable machine learning models (using SHAP) to AWS cloud infrastructure.

Three approaches are shown:

  1. Using Sagemaker. This involves setting up custom training containers on ECR, and defining the proper inference functions.
    • All notebooks, code, configurations, Dockerfiles and READMEs can be found in the sagemaker folder.
  2. Using straight AWS Lambda functions and zappa. This is easier, but still a number of tricky things to get to work (such as deploying from inside a lambda compatible docker container)
    • All code, Makefiles and READMEs can be found in the lambda folder.
  3. For completeness an example of a local on premise deployment can be in the local folder.

An example dashboard that sends requests to both a sagemaker deployment and a lambda deployment is running at http://creditexplainer.herokuapp.com

About

Example of how to deploy an ML algorithm together with SHAP explanations to AWS Sagemaker, including a front end dashboard.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published