Skip to content

Latest commit

 

History

History
35 lines (20 loc) · 795 Bytes

File metadata and controls

35 lines (20 loc) · 795 Bytes

ConvReLU++: Reference-based Lossless Acceleration of Conv-ReLU Operations for Mobile Deep Vision

ConvReLU++ is a lossless fast Conv-ReLU kernel without any model training. This repo contains official PyTorch implementation of ConvReLU++.

Setup

pip install torch torchvision Pillow

Usage example

Citation

@inproceedings{kong2023convrelu++,
  title={Convrelu++: Reference-based lossless acceleration of conv-relu operations on mobile cpu},
  author={Kong, Rui and Li, Yuanchun and Yuan, Yizhen and Kong, Linghe},
  booktitle={Proceedings of the 21st Annual International Conference on Mobile Systems, Applications and Services},
  pages={503--515},
  year={2023}
}