Requirements and check methods
- Ubuntu 14.04 or higher
- CUDA 10.1 or higher
nvcc --version - pytorch 1.3 or higher
python import torch torch.__version__ - python 3.6 or higher
- gcc-7 or higher
gcc -v g++ -v
Installation
The Minkowski engine github page gives three ways to install: pip, anaconda, or on the system directly. Here I strongly recommand installing it using anaconda since our servers are managed based on anaconda environment.
1. Create a conda virtual environment and install requirements.
conda create -n py3-mink python=3.7
conda activate py3-mink
conda install numpy mkl-include
conda install pytorch=1.4.0 -c pytorch
Here is a trick when you meet issues like installing packages using conda too slowly (click here for details).
First, open your condarc
vim ~/.condarc
Next modify it like this
channels:
- https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
- https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
- https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/
- https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/msys2/
- default
ssl_verify: true
show_channel_urls: true
Then save and exit, source it again
source ~/.condarc
After these, conda install again, but note that don’t add commands like ‘-c pytorch’ because ‘-c’ is an option for a download channel. If conda install without ‘-c’, it just goes to ~/.condarc to search for download websites.
2. Compilation and installation
conda activate py3-mink
git clone https://github.com/StanfordVL/MinkowskiEngine.git
cd MinkowskiEngine
python setup.py install
3. Check installation
python
import MinkowskiEngine as ME
ME.__version__