To start with, clone the SDANNCE repository to your local machine, if not already:
git clone https://github.com/tqxli/sdannce.git
cd sdannce
Prepare the runtime environment using Conda:
conda create --name sdannce python=3.8
conda activate sdannce
# REPLACEABLE with other workable PyTorch installation
conda install pytorch=1.9.1 torchvision=0.10.1 cudatoolkit=11.1 cudnn ffmpeg -c pytorch -c nvidia
pip install setuptools==59.5.0
pip install -e .
The installation was tested on Linux (Ubuntu 16.04, 20.04) with a NVIDIA GPU (Titan V, RTX 3090, A5000, A6000) and Conda version=4.10.3.
The code was tested for Python 3.7-3.8, PyTorch 1.8-1.9.1 (and torchvision that matches the PyTorch installation according to the official instructions https://pytorch.org/get-started/previous-versions/).
If you encounter any issues with installing PyTorch (at the step conda install pytorch=1.9.1 ...
), specifically, using modern versions of Conda, please install with pip instead:
pip install torch==1.9.1+cu111 torchvision==0.10.1+cu111 -f https://download.pytorch.org/whl/torch_stable.html
Users should be mindful of choosing PyTorch versions compatible with their local CUDA installation.