Dolphin

Dolphin is a machine learning platform built on top of Apache REEF. Dolphin consists of a BSP-style machine learning framework (dolphin-bsp), a deep learning framework (dolphin-dnn), and a parameter server module (dolphin-ps).

Submodules

• dolphin-bsp: BSP-style framework for multi-staged processing.
• dolphin-dnn: Deep learning framework for training deep neural network models on large datasets.
• dolphin-ps: Parameter server module for asynchronous machine learning algorithms.

How to build and run Dolphin

1. Build REEF: check https://cwiki.apache.org/confluence/display/REEF/Compiling+REEF Currently, Dolphin depends on REEF 0.14.0-SNAPSHOT, which means you must build the current snapshot of REEF before building Dolphin. We will move to the release version 0.14.0 once it's out.

2. Build Dolphin:

   git clone https://github.com/cmssnu/dolphin
   cd dolphin
   mvn clean install
   

3. Run Dolphin: check the READMEs in the submodules for more details.

Dolphin Mailing List

Bug reports and feature requests are welcome, as well as even simple questions!
Contact us and share your thoughts by subscribing to dolphin-discussion@googlegroups.com.