You are here: Start » PyAVL » Installation

Installation

PyAVL and PyAVL-DL are installed with pip. Both packages are provided as directories inside the Aurora Vision Library installation, and pip installs them directly from these directories. Python 3.12 or later is required.

Installation Files

Each package and all its supporting files are placed in a single directory. The directory contains everything that pip needs to build and register a wheel.

PyAVL

AVL.pyd on Windows, AVL.so on Linux —the PyAVL extension module.
AVL.pyi —the type stub file for code completion and static type checking in editors.
pyproject.toml and hatch_build.py —the build metadata and the build hook, which packages the native files into a wheel.
AVL.dll and its runtime dependencies on Windows, or the shared libraries on which the module depends on Linux.

PyAVL-DL

AVLDL.pyd on Windows, AVLDL.so on Linux —the PyAVL-DL extension module.
AVLDL.pyi —the type stub file.
pyproject.toml and hatch_build.py —the build metadata and the build hook.
AVLDL.dll and its runtime dependencies on Windows, or the shared libraries on which the module depends on Linux.
On Windows, the CUDA and ONNX Runtime libraries from the Deps_x64 directory (GPU builds).

No files need to be copied or moved.

Installing on Windows

The Aurora Vision Library installer sets the environment variable AVL_PATH5_7 to the installation directory. The Aurora Deep Learning installer sets AVLDL_PATH5_7 in the same way. The packages are located in the bin\x64 subdirectory of these directories.

Open a command prompt window.
Install PyAVL:
```
pip install "%AVL_PATH5_7%\bin\x64"
```
Install PyAVL-DL. This step is optional and requires the Aurora Deep Learning add-on. PyAVL-DL depends on PyAVL, so PyAVL has to be installed first.
```
pip install "%AVLDL_PATH5_7%\bin\x64"
```

After the installation, the modules can be imported in every Python session that uses this Python environment.

Installing on Linux

Unpack the library archive and run the install script, as described in Using Library on Linux. The script installs the library to the directory /opt/AuroraVision/AVL-5.7.<Build>. The Deep Learning archive is installed in the same way to /opt/AuroraVision/AVLDL-5.7.<Build>.

The packages are located in the lib/<architecture> subdirectory, where <architecture> is x86_64-linux-gnu or aarch64-linux-gnu, depending on the system. Install PyAVL first and then, optionally, PyAVL-DL:

pip install "/opt/AuroraVision/AVL-5.7.<Build>/lib/x86_64-linux-gnu/"
pip install "/opt/AuroraVision/AVLDL-5.7.<Build>/lib/x86_64-linux-gnu/"

The build hook packages the extension module, the type stubs and all the shared libraries into the wheel, so it is not necessary to configure LD_LIBRARY_PATH.

Verifying the Installation

Import the installed modules. The command on Linux uses python3 instead of python:

python -c "import AVL; print('AVL imported successfully')"
python -c "import AVLDL; print('AVLDL imported successfully')"

The installed packages are named pyavl and pyavl-dl, and pip show pyavl displays the installed version.

Updating and Removing

To update, install the package again from the directory of the newer Aurora Vision Library installation. To remove the packages, uninstall PyAVL-DL first and then PyAVL:

pip uninstall pyavl-dl pyavl