Welcome to Rafiki’s Documentation!

What is Rafiki?

Rafiki is a distributed system that trains machine learning (ML) models and deploys trained models, built with ease-of-use in mind. To do so, it leverages on automated machine learning (AutoML).

For Application Developers and Application Users, without any ML expertise, they can:

• Create a model training job for supported tasks, with their own datasets

• Deploy an ensemble of trained models for inference

• Integrate model predictions in their apps over HTTP

For Model Developers, they can:

• Contribute to Rafiki’s pool of model templates

Check out Quick Setup to deploy/develop Rafiki on your machine, and/or Quick Start to use a deployed instance of Rafiki.

Index

• User Guide
  • Quick Start
  • Quick Start (Model Developers)
  • Quick Start (Application Developers)
  • Quick Start (Application Users)
  • Quick Start (Admins)
  • Using Rafiki’s Web Admin
  • Supported Tasks
  • Dataset Types
  • Installing Python
  • Model Development Guide
  • How Model Tuning Works
• Developer Guide
  • Setup & Configuration
  • Development
  • Users of Rafiki
  • Rafiki’s Architecture
  • Folder Structure
• Python Documentation
  • rafiki.client
  • rafiki.model
  • rafiki.constants

Issues

Report any issues at Apache SINGA’s JIRA or Rafiki’s Github Issues.

Acknowledgements

The research is supported by the National Research Foundation, Prime Minister’s Office, Singapore under its National Cybersecurity R&D Programme (Grant No. NRF2016NCR-NCR002-020), National Natural Science Foundation of China (No. 61832001), National Key Research and Development Program of China (No. 2017YFB1201001), China Thousand Talents Program for Young Professionals (3070011 181811).