~ 2016 2Q
- Hybrid (sync/async) distributed deep learning on CPU cluster
- Release 0.1 version
- BSP based mini-batch SGD
- Programming Interface
- Scalability and Performance Evaluationcompare with CPU only Caffe, Singa, DeepDist, ..., etc.
~ 2016 4Q
...
- CPU + GPU heterogeneous cluster optimization and Evaluation
- Release 0.2 version
- If possible, Integrate with Caffe, Singa, Torch, ..., etc
- Performance Evaluation
Overview
Content Tools
ThemeBuilder
Apps