CNN: Caffe on Petuum
Caffe is a deep learning framework originally developed by Yangqing Jia / BVLC (http://caffe.berkeleyvision.org/) where GPU was used to speedup the model training. Here we provide a CPU distributed version of Caffe based on Petuum. Good speedup and scalability is achieved. We have kept most of the Caffe interfaces unchanged.
Please consult the wiki page for more details.