Accelerator-aware Neural Network Design using AutoML

Abstract

While neural network hardware accelerators provide a substantial amount of raw compute throughput, the models deployed on them must be co-designed for the underlying hardware architecture to obtain the optimal system performance. We present a class of computer vision models designed using hardware-aware neural architecture search and customized to run on the Edge TPU, Google's neural network hardware accelerator for low-power, edge devices. For the Edge TPU in Coral devices, these models enable real-time image classification performance while achieving accuracy typically seen only with larger, compute-heavy models running in data centers. On Pixel 4's Edge TPU, these models improve the accuracy-latency tradeoff over existing SoTA mobile models.

arXiv commentAccepted paper at the On-device Intelligence Workshop at MLSys Conference 2020

References in corpus (1)

• MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications

Cited by in corpus (11)

• EfficientNetV2: Smaller Models and Faster Training
• A Comprehensive Survey on Hardware-Aware Neural Architecture Search
• MobileDets: Searching for Object Detection Architectures for Mobile Accelerators
• Weight-Sharing Neural Architecture Search: A Battle to Shrink the Optimization Gap
• Evaluating Pre-trained Convolutional Neural Networks and Foundation Models as Feature Extractors for Content-based Medical Image Retrieval
• Rethinking Co-design of Neural Architectures and Hardware Accelerators
• Apollo: Transferable Architecture Exploration
• S3NAS: Fast NPU-aware Neural Architecture Search Methodology
• Discovering Multi-Hardware Mobile Models via Architecture Search
• Data-Driven Offline Optimization For Architecting Hardware Accelerators
• ISyNet: Convolutional Neural Networks design for AI accelerator