If your distribution has at least Python 3.7, and you just wish to separate tracks with Demucs, not train it, you can just run
pip3 install --user -U demucs
# Then anytime you want to use demucs, just do
python3 -m demucs -d cpu PATH_TO_AUDIO_FILE_1
# If you have added the user specific pip bin/ folder to your path, you can also do
demucs -d cpu PATH_TO_AUDIO_FILE_1If Python is too old, or you want to be able to train, I recommend installing Miniconda, with Python 3.7 or more.
conda activate
python3 install -U demucs
# Then anytime you want to use demucs, first do conda activate, then
demucs -d cpu PATH_TO_AUDIO_FILE_1Of course, you can also use a specific env for Demucs.