Install HOOPS AI

HOOPS AI Package & License Key

HOOPS AI comes with a 60-day evaluation license. During your evaluation, a dedicated Consulting Engineer will be assigned to assist you. For technical support or to extend your evaluation license, please contact your assigned Consulting Engineer.

Tip

Start your Trial Experience to get immediate access to HOOPS AI.

Python Version Requirement

HOOPS AI requires Python 3.12 exactly. Python 3.11 and earlier are not supported, and Python 3.13 and later are not supported.

Check your version:

python --version

Download Python 3.12 from python.org if needed.

Installation Methods

Pick the method that fits your setup:

  • Install with pip – standard Python virtual environment using pip. Good if you already have Python 3.12 available.

  • Install with uv – uv manages the Python version and virtual environment for you. Recommended if you are starting fresh.

Both methods install from the same public feed and produce an equivalent environment.

Note

If you have Conda installed and pip or python still point to a Conda-managed path after activating your virtual environment, see Removing Conda.

Extras

Install hoops-ai[all] to get the full stack in one command, including CAD reading, ML libraries, visualization, format conversion, the web viewer, and notebook utilities.

The full dependency list is on the Acknowledgments page.

GPU and CPU Support

HOOPS AI is the same package whether you run on CPU or GPU. The only difference at install time is which PyTorch index you point pip to.

The recommended GPU configuration uses PyTorch 2.9 with CUDA 13.0 (index tag cu130). PyTorch bundles the CUDA runtime, so you do not need to install the CUDA toolkit separately. If you need a different CUDA version, the PyTorch installation page lists all available index tags for each PyTorch release.

You do need a recent NVIDIA driver. For CUDA 13.0, driver version 575.51 or newer is required. Run nvidia-smi to check yours.

See Install with pip or Install with uv for the exact install commands.

Registering the Jupyter Kernel

If you plan to use HOOPS AI in JupyterLab or Jupyter Notebook, register your virtual environment as a kernel after installation. This requires ipykernel, which is included in the notebooks and all extras.

.venv/bin/python -m ipykernel install --user \
    --name my-project \
    --display-name "My Project"

Replace my-project and "My Project" with a name that identifies your project. The --name is used internally and must be unique on your system; the --display-name is what appears in the JupyterLab kernel selector.

After registering, open JupyterLab, click the kernel selector in the top-right corner, choose Select another kernel, then Jupyter Kernel, and pick your kernel from the list. Click the refresh button if it does not appear immediately.

In VS Code, open the notebook, click the kernel selector in the top-right corner, choose Select another kernel, then Jupyter Kernel, and pick your kernel. Click the refresh button if it does not appear.

Troubleshooting

“CUDA available: False” after GPU installation

If torch.cuda.is_available() returns False:

  • Your NVIDIA driver may be outdated. Driver 575.51 or newer is required for CUDA 13.0.

  • The GPU may not be CUDA-compatible. Check that its compute capability is 3.5 or higher.

Verify your driver:

nvidia-smi

“torch is not a GPU build”

If you installed without the PyTorch index flag, you may have ended up with a CPU-only build. Reinstall with the correct index:

pip install "torch>=2.9.0,<2.10.0" \
    --extra-index-url https://download.pytorch.org/whl/cu130

Import errors

If import hoops_ai fails, check that your virtual environment is activated and that you installed into the right one:

which python
python -c "import hoops_ai; print(hoops_ai.__version__)"

pip points to the wrong location

If which pip (Linux) or where pip (Windows) points to a system or Conda path instead of your virtual environment, see Removing Conda.

“Could not find a version that satisfies the requirement”

This usually means pip is running under the wrong Python version. HOOPS AI requires Python 3.12 exactly.

Check which Python your environment is using:

python --version
which python

If the version is not 3.12, recreate your virtual environment with Python 3.12. See Install with pip or Install with uv for the steps.

Getting Started with Tutorials

Tutorial Notebooks

Access our interactive Jupyter notebooks from the public GitHub repository:

Clone the repository or download the notebooks directly to get started with step-by-step examples.

You can also find the tutorials in this documentation: Tutorials

Tutorial Data Assets

The tutorials require sample CAD files and datasets. Download the tutorial data package using the link provided in your Sales team email, or access it directly:

After downloading, extract the data to a convenient location and update the file paths in the tutorial notebooks accordingly.

Note

Make sure to activate your virtual environment and select the HOOPS AI Jupyter kernel before running the tutorials.

Next Steps

Now that you have HOOPS AI installed and know where to get an extended evaluation or perpetual license key, you are ready to read the following pages:

Ready? Simply press the button Next below to get started!