An Once-for-All Budgeted Pruning Framework for ConvNets Considering Input Resolution
This work provides a more efficient method for deploying deep learning models on edge devices with dynamic computational budgets, which is a problem for practitioners and researchers in mobile AI.
This paper proposes OFARPruning, a framework for finding compact convolutional neural network structures that adapt to various input resolutions and FLOPs constraints with a single training. It achieves 1-2% higher accuracy than existing once-for-all compression methods and matches or exceeds conventional pruning methods (e.g., 72.6% vs. 70.5% on MobileNetv2 under 170 MFLOPs) with significantly higher efficiency.
We propose an efficient once-for-all budgeted pruning framework (OFARPruning) to find many compact network structures close to winner tickets in the early training stage considering the effect of input resolution during the pruning process. In structure searching stage, we utilize cosine similarity to measure the similarity of the pruning mask to get high-quality network structures with low energy and time consumption. After structure searching stage, our proposed method randomly sample the compact structures with different pruning rates and input resolution to achieve joint optimization. Ultimately, we can obtain a cohort of compact networks adaptive to various resolution to meet dynamic FLOPs constraints on different edge devices with only once training. The experiments based on image classification and object detection show that OFARPruning has a higher accuracy than the once-for-all compression methods such as US-Net and MutualNet (1-2% better with less FLOPs), and achieve the same even higher accuracy as the conventional pruning methods (72.6% vs. 70.5% on MobileNetv2 under 170 MFLOPs) with much higher efficiency.