Skip to content

Latest commit

 

History

History
35 lines (34 loc) · 1.04 KB

INSTALL.md

File metadata and controls

35 lines (34 loc) · 1.04 KB

Prerequisites

  • Lnux
  • Python3
  • NVIDIA GPUs + CUDA CuDNN

Preparation

You can set up the environments in your own way (requirements.txt has been provided), or following below steps:

  1. Clone this repo with: 
    git clone https://github.com/HobbitLong/PyContrast.git
cd PyContrast/pycontrast
  2. (Optional) you can consider setting up a virtual environment. Such environment (e.g. PyConEnv) can be created by 
    virtualenv -p python3 ~/env/PyConEnv
    then activate it: 
    source ~/env/PyConEnv/bin/activate
  3. Install packages: 
    pip3 install -r requirements.txt
  4. (Optional) install apex if you would like to try mixed precision training. If you do not want to take a look at the apex repo, the installing commands are (assuming pytorch and CUDA are availabel): 
    cd ~
git clone https://github.com/NVIDIA/apex
cd apex
pip install -v --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./