CASL’s paper accepted to PACT 2019
A research paper entitled “MOSAIC: Heterogeneity-, Communication-, and Constraint-Aware Model Slicing and Execution for Accurate and Efficient Inference” has been accepted for publication at PACT’19. The paper is co-authored by Myeonggyun Han, Jihoon Hyun, Seongbeom Park, Jinsu Park, and Prof. Woongki Baek at Computer Architecture and Systems Lab. (CASL), CSE, UNIST.
This work proposes MOSAIC, heterogeneity-, communication-, and constraint-aware model slicing and execution for accurate and efficient inference on heterogeneous embedded systems. MOSAIC generates the efficient model slicing and execution plan for the target inference workload through dynamic programming. MOSAIC significantly reduces inference latency and energy, exhibits high estimation accuracy, and incurs small overheads.
PACT (International Conference on Parallel Architectures and Compilation Techniques) is one of the top-tier conferences in the field of parallel computing.