OverSeer: Efficient DNN Execution on Heterogeneous Devices

Deep learning applications have become pervasive in our daily lives, and are being executed on awide variety of heterogeneous devices. This paper presents OverSeer , a topologyand hardware-agnostic end-to-end system, that uses a novel partitioning algorithm to efficiently utilize different computational capabilities present in heterogeneous devices. ACM Reference Format: Rohan Mukherjee1,2, Zhi Chen1, Vin Sharma1, Animesh Jain1. 2019. OverSeer: Efficient DNN Execution on Heterogeneous Devices. In Proceedings of ACM Conference (Workshop on AI Systems at SOSP’19). ACM, New York, NY, USA, 3 pages. https://doi.org/10. 1145/nnnnnnn.nnnnnnn