A flexible and efficient deep neural network (DNN) compiler that generates high-performance executable from a DNN model description.

Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels