Installing Demucs Locally?
See original GitHub issue❓ Questions
I genuinely have no idea what I’m doing wrong. I don’t have any Python experience, I just want to be able to run Demucs locally (I do have a GPU). Off the top of my head:
- I’ve installed python
- I’ve installed the anaconda prompt
- I’ve installed the github desktop client
- I’ve downloaded/installed pytorch
- I’ve run
python.exe -m pip install -U demucs
However, none of the following from the Windows support page work:
cd %HOMEPATH%
git clone -b main --single-branch https://github.com/facebookresearch/demucs ./demucs
“‘git’ is not recognized as an internal or external command, operable program or batch file.”
cd ./demucs
“The system cannot find the path specified.”
conda env update -f environment-cpu.yml
EnvironmentFileNotFound: ‘C:\Users\Upstairs\environment-cpu.yml’ file not found
conda activate demucs
“Could not find conda environment: demucs. You can list all discoverable environments with conda info --envs
.”
python.exe -m pip install -e .
“ERROR: File “setup.py” or “setup.cfg” not found. Directory cannot be installed in editable mode: C:\Users\Upstairs”
What am I doing wrong? Is there something I need to do to get Github to interface with the anaconda prompt?
Issue Analytics
- State:
- Created 2 years ago
- Comments:8 (1 by maintainers)
Top GitHub Comments
Hey, here’s a beginner’s guide to installing demucs locally from source:
Step 1 Open Anaconda Prompt
Step 2 Create a conda environment for demucs using:
conda create --name demucs
Type
y
when prompted to proceed.Step 3 Activate the environment with:
conda activate demucs
Step 4 Install Pytorch for CUDA using:
conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch
Step 5 Since you’re only interested in running the code, no need to bother with the
git
command or GitHub desktop. Just download and extract thedemucs-main
zip from theCode
dropdown on the main page:Step 6 Install
demucs
by entering the following in the anaconda prompt window:pip install -e "PATH_TO_demucs-main"
Where PATH_TO_demucs-main is your path to the extracted
demucs-main
folder, so the final command might look like:pip install -e "C:\Users\username\Documents\demucs-main"
Step 7 Install PySoundFile audio backend using:
pip install PySoundFile
That’s it, you’re all setup!
Step 8 Test if everything is running using:
demucs -d cuda -o PATH_TO_OUTPUT_DIRECTORY PATH_TO_AUDIO_FILE
Now, whenever you want to run demucs:
demucs
command - Step 8Note: Usually you would use the .yml file to automatically create the environment and download all dependencies, but for some reason,
environment-cuda.yml
kept downloading the cpu version of pytorch, so I had to create the environment and install pytorch manually.So, it turns out I was massively overcomplicating things! Since I had already run
python.exe -m pip install -U demucs
andpip install PySoundFile
, all I needed to do was give it something to split - the Windows support page is irrelevant. Maybe this could be made clearer in the readme?(Thanks again for your help, thesinepainter!)