docker build -t cuda-env .
docker run --gpus all -it cuda-env
docker run --gpus all -it -v 'D:\EarlySteps\capsule-6073072:/workspace' cuda-env
---------------------------
next steps
python start.py , you get some errors which can be fixed by these cmds
cd /workspace/code/torchlight
python setup.py install
cd ..
pip install --upgrade numpy
pip install "numpy>=1.20"