Once Quantization-Aware Training: High Performance
             Extremely Low-bit Architecture Search
       Mingzhu Shen1 Feng Liang1 Ruihao Gong1 Yuhang Li1 Chuming Li1
           Chen Lin2 Fengwei Yu1 Junjie Yan1 Wanli Ouyang3
        1
         Sensetime Research 2 University of Oxford 3 The University of Sydney
     shenmingzhu@sensetime.com, liangfeng@sensetime.com, wanli.ouyang@sydney.edu.au

            Abstract
  Quantization Neural Networks (QNN) have attracted a
lot of attention due t ...