DaeHwanGi / ActiveShiftLayer_pytorch

pytorch implementation Active Shift Layer

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Active Shift Layer

pytorch Constructing Fast Network through Deconstruction of Convolution implementation

It does not support CUDA now.

Prerequisite

  • Windows10
  • Pytorch 1.1.0
  • cuda 9.0

Build

???@???:~$ python ./DepthwiseAffineGrid/setup.py install
???@???:~$ python ./DepthwiseGridSample/setup.py install

To DO

  • use C++ extension
  • Write Backward
  • Support CUDA tensor
  • Comparison with paper
  • Improve performance using CUDA

About

pytorch implementation Active Shift Layer


Languages

Language:C++ 53.3%Language:Cuda 24.5%Language:Python 22.2%