Tensor Processing Primitives: A Programming Abstraction for Efficiency and Portability in Deep Learning Workloads

Open in new window