First the venv must be created and dependencies installed. The Makefile
has a good start on how to do so. Running make should work for
Debian based systems. It clones Parla.py and installs it using its setup.py
There is a manual part related to llvm, which is printed during the make.
After the venv is created, run source ./activate.sh to activate it.
container dir contains a future container for this workload due to machines having different CUDA versions, it's not working at this point.