Distilling Optimal Neural Networks: Rapid Search in Diverse Spaces
This work provides a significantly faster and more versatile method for Neural Architecture Search, benefiting researchers and practitioners who need to design efficient neural networks for various hardware platforms and diverse architectural constraints.
This paper introduces DONNA, a novel pipeline for Neural Architecture Search (NAS) that addresses the limitations of current methods in scaling to multiple hardware platforms and handling diverse architectural search-spaces. DONNA achieves up to 100x faster search than MNasNet, and its architectures are 20% faster than EfficientNet-B0 and MobileNetV2 on a Nvidia V100 GPU for ImageNet classification, and 10% faster with 0.5% higher accuracy than MobileNetV2-1.4x on a Samsung S20 smartphone.
Current state-of-the-art Neural Architecture Search (NAS) methods neither efficiently scale to multiple hardware platforms, nor handle diverse architectural search-spaces. To remedy this, we present DONNA (Distilling Optimal Neural Network Architectures), a novel pipeline for rapid, scalable and diverse NAS, that scales to many user scenarios. DONNA consists of three phases. First, an accuracy predictor is built using blockwise knowledge distillation from a reference model. This predictor enables searching across diverse networks with varying macro-architectural parameters such as layer types and attention mechanisms, as well as across micro-architectural parameters such as block repeats and expansion rates. Second, a rapid evolutionary search finds a set of pareto-optimal architectures for any scenario using the accuracy predictor and on-device measurements. Third, optimal models are quickly finetuned to training-from-scratch accuracy. DONNA is up to 100x faster than MNasNet in finding state-of-the-art architectures on-device. Classifying ImageNet, DONNA architectures are 20% faster than EfficientNet-B0 and MobileNetV2 on a Nvidia V100 GPU and 10% faster with 0.5% higher accuracy than MobileNetV2-1.4x on a Samsung S20 smartphone. In addition to NAS, DONNA is used for search-space extension and exploration, as well as hardware-aware model compression.