PLiNIO: A User-Friendly Library of Gradient-based Methods for Complexity-aware DNN Optimization
It addresses the need for efficient DNNs on constrained edge devices, but it is incremental as it integrates existing optimization techniques into a unified library.
The paper tackles the problem of automating DNN optimization for edge devices by proposing PLiNIO, a library that combines gradient-based methods, resulting in up to 94.34% memory reduction with less than 1% accuracy drop compared to baselines.
Accurate yet efficient Deep Neural Networks (DNNs) are in high demand, especially for applications that require their execution on constrained edge devices. Finding such DNNs in a reasonable time for new applications requires automated optimization pipelines since the huge space of hyper-parameter combinations is impossible to explore extensively by hand. In this work, we propose PLiNIO, an open-source library implementing a comprehensive set of state-of-the-art DNN design automation techniques, all based on lightweight gradient-based optimization, under a unified and user-friendly interface. With experiments on several edge-relevant tasks, we show that combining the various optimizations available in PLiNIO leads to rich sets of solutions that Pareto-dominate the considered baselines in terms of accuracy vs model size. Noteworthy, PLiNIO achieves up to 94.34% memory reduction for a <1% accuracy drop compared to a baseline architecture.