Energy-Aware DNN Graph Optimization
This addresses energy consumption issues for resource-constrained machine learning devices, representing an incremental improvement over existing performance-focused optimization methods.
The paper tackles the problem of optimizing deep neural network graphs for energy efficiency on power-constrained devices, achieving 24% energy savings with negligible performance impact.
Unlike existing work in deep neural network (DNN) graphs optimization for inference performance, we explore DNN graph optimization for energy awareness and savings for power- and resource-constrained machine learning devices. We present a method that allows users to optimize energy consumption or balance between energy and inference performance for DNN graphs. This method efficiently searches through the space of equivalent graphs, and identifies a graph and the corresponding algorithms that incur the least cost in execution. We implement the method and evaluate it with multiple DNN models on a GPU-based machine. Results show that our method achieves significant energy savings, i.e., 24% with negligible performance impact.