Installation failure of `torch-points-kernels` due to missing PyTorch in venv

[arnavk23]

I am trying to install the dependencies for RandLA-Net, specifically the torch-points-kernels package, inside a Python virtual environment (venv). However, the installation fails with the error:

ModuleNotFoundError: Please install pytorch >= 1.1 before proceeding.

even though PyTorch is installed in the same venv prior to running the install.

Steps to reproduce:

1. Create and activate a Python 3.12 virtual environment:

python3.12 -m venv venv
source venv/bin/activate

2. Upgrade pip and install PyTorch:

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

3. Attempt to install torch-points-kernels:

pip install torch-points-kernels -f https://data.pyg.org/whl/torch-2.0.1+cu118.html

4. See error output:

ModuleNotFoundError: Please install pytorch >= 1.1 before proceeding.

Additional information:

• Python version: 3.12.0
• PyTorch version (inside venv): Confirmed installed and importable via python -c "import torch; print(torch.__version__)"
• OS: Ubuntu 24.04
• Installation inside a clean venv
• The error occurs during the build stage when pip tries to get requirements for the wheel.

Workarounds tried:

• Verified PyTorch is installed and importable in the venv before installing torch-points-kernels.
• Tried reinstalling PyTorch.
• Attempted installation on Python 3.10, which may have better compatibility.
• Created fresh venv to isolate the problem.
• Still fails with the same error about missing PyTorch.

Request:

Please advise on how to resolve this installation error for torch-points-kernels. It seems like the package’s build process does not detect PyTorch installed in the current environment, which blocks the installation.

Thank you!