Requirements:

python 2.7 tensorflow 1.13.1 ImageNet dataset

For training and test:

python imagenet_main.py

This is an implementation example of the complete quantization framework named WAGEUBN proposed by Yukuan Yang, etc. If you have any questions, please feel free to contact me. My email is yyk17@mails.tsinghua.edu.cn. For more details, please refer to my paper.

Citation:

If you find this paper or this repository helpful, please cite it:

@article{YANG202070,
title = "Training high-performance and large-scale deep neural networks with full 8-bit integers",
journal = "Neural Networks",
volume = "125",
pages = "70 - 82",
year = "2020",
issn = "0893-6080",
author = "Yukuan Yang and Lei Deng and Shuang Wu and Tianyi Yan and Yuan Xie and Guoqi Li"
}