Fix for linux installation

[mithusingh32]

• Updated numpy version to 1.24 or higher
• Removed Dora references
• Update README

[ChrisTG742]

@mithusingh32 did you test that updating numpy does not cause incompatibilites? There was/is some problems with librosa and numpy: #1016
Or did you update librosa aswell?

[mithusingh32]

I did not update librosa.

I did some initially testing w/ demucs v4 htddemucs_ft with all default parameters and it worked.
I will try perform more testing with other models + parameters.

GPU: RTX 3070ti 8GB

[shervinemami]

Hi @mithusingh32 , I tried your Linux fix and it worked for me on Kubuntu 24.04. I tested with htdemucs_ft model on GPU as well as MDXNet using kuielab model on GPU. I had to rollback to using the older Python 3.11 instead of Python 3.12 that my system comes with.

I recommend adding this info to the Linux installation README so that all users can use the same Python version no matter their system comes with:

# Install UVR5 into a Python 3.11 venv (not virtualenv):
sudo add-apt-repository ppa:deadsnakes/ppa   # Access many Python versions
sudo apt update         # Allow installing many Python versions
sudo apt install python3.11        # Install Python 3.11
sudo apt install python3.11-venv   # So you can create venv virtual environments
sudo apt install python3.11-dev    # So you can build your own python packages, instead of only downloading them off the web
sudo apt install python3.11-tk     # So UVR can display graphics
python3.11 -m venv py3.11_UVR5.venv    # Create a new venv using Python 3.11
source py3.11_UVR5.venv/bin/activate
pip3 install --upgrade pip wheel   # Install or update the pip & wheel versions that are inside the venv
SKLEARN_ALLOW_DEPRECATED_SKLEARN_PACKAGE_INSTALL=True pip3 install -r requirements.txt

# Run UVR5 in the Python virtual environment:
source py3.11_UVR5.venv/bin/activate
python3 UVR.py