snuspl / cruise

Cruise: A Distributed Machine Learning Framework with Automatic System Configuration

Cruise

Cruise is a distributed machine learning (ML) framework with automatic system configuration, built on top of Apache REEF. Currently, Cruise consists of a Parameter Server (Cruise PS) for asynchronous ML training and a BSP-style engine (Cruise Pregel) for processing data that we can represent as a graph.

Submodules

• ps: An asynchronous ML processing engine based on Parameter Server.
• pregel: A BSP-style graph processing engines
• elastic-tables: An Elastic Runtime that allows changing system configurations transparently at runtime.

Requirements

• Java 8 JDK
• Maven 3
• $ sudo apt-get install libgfortran3 (Ubuntu)

How to build

git clone https://github.com/snuspl/cruise
cd cruise
mvn clean install

How to run

Please take a look at the instructions for submodules (ps and pregel)

Mailing list

All type of discussions such as questions, bug reports, and feature requests are always welcome! Please contact us and share your thoughts by subscribing to [e-mail]