requirements-test.txt

pytest>=6.2.4
pytest-cov>=2.12.1

# requires pytorch to be installed first (duplicated in requirements-experiment.txt)
# - we need `nvcc` to be installed first, otherwise GPU kernel extensions will not be
#   compiled and this error will silently be skipped. If you get an error such as
#   $ conda install -c nvidia cuda-nvcc
# - Make sure that the version of torch corresponds to the version of `nvcc`, torch needs
#   to be compiled with the same version! Install the correct version from:
#   https://pytorch.org/get-started/locally/ By default torch compiled with 10.2 is installed,
#   but `nvcc` will probably want to install 11.
#   CUDA 10.2 (as of 2022-03-15) EITHER OF:
#       $ pip3 install torch torchvision torchaudio
#       $ conda install pytorch torchvision torchaudio cudatoolkit=10.2 -c pytorch
#   CUDA 11.3 (as of 2022-03-15) EITHER OF:
#       $ pip3 install torch==1.11.0+cu113 torchvision==0.12.0+cu113 torchaudio==0.11.0+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html
#       $ conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch
# - I personally just manage my cuda version manually, installing the correct cudatoolkit from: https://developer.nvidia.com/cuda-toolkit-archive
#   Then making sure that:
#       PATH contains:            "/usr/local/cuda/bin"
#       LD_LIBRARY_PATH contains: "/usr/local/cuda/lib64"
torchsort>=0.1.4