The focus of this project is to understand the performance of different machine learning frameworks specifically the ones that have implemented deep learning and neural networks on KNL systems. At first two frameworks, Caffe2 and Pytorch will be examined. A flat profile will be determined and different parameters such as batch size and epochs numbers will be used to determine the time to solution and also accuracy.

These frameworks are written in C/C++ files, while their API runs on Python 2. The performance then will be expanded on different data sets and types.