To start, at least ensure these packages are installed.
torch==2.0.0
setuptools
This code is tested on cuda-11.7 / cuda-11.8, with gcc-9 and g++-9 / gcc-11 and g++-11, if you are using conda environments, this should work. And remember to link your cuda in the environment.
conda install cudatoolkit=11.7
conda env create -f environment.yaml
- (optional) if you are using conda, activate an environment that you want to work with
- If you just want to compile the extension, use
python setup.py build
to build the extension. Uppon sussessfully compilation, you can find the*.so
file inbuild/lib.*/
. You can add this path to your python path orcd
to this directory, and then start ipython to import the module - For installation you can run
python setup.py install
, and the extension will be installed, and then you can access it system-wise.
This is the updated version (torch 2.0.0, cuda 11.7, gcc-9, g++-9) of torch_bingham, which is built upon Bingham Statistics Library (BSL).