GPU testing / parallelization

Virtual environment

Use a local environment in .venv under this project directory so CUDA wheels and version pins do not affect your global Python.

Create (from this folder, in PowerShell or Command Prompt):

python -m venv .venv

Or double-click or run setup_venv.bat in this folder; it creates .venv and, on a fresh create only, runs the same pip install commands as in Dependencies below. If .venv already exists, the script exits without reinstalling packages.

Activate (do this before running your code or any extra pip install):

• PowerShell: .venv\Scripts\Activate.ps1
• Command Prompt: .venv\Scripts\activate.bat

If PowerShell blocks the script, use Command Prompt activation or run Set-ExecutionPolicy -Scope CurrentUser RemoteSigned once.

Dependencies (old Windows + GPU)

Specific environment required for GPU use on an older Windows setup.

PyTorch (CUDA 11.8 wheels):

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

NumPy (stay below 2.x for compatibility):

pip install "numpy<2"

scikit-learn 1.8:

pip install scikit-learn==1.8