Python interface
Installation of the Python package
The Python package is stored as a whl file in the folder <TwinCatInstallDir>\Functions\TF38xx-Machine-Learning\Utilities\ModelManagerAPI\PythonPackage.
The package is compatible with:
- Windows 10
- Windows 11
- Debian 12
- Ubuntu 24.04
The Clang C++ libraries are required for use under Linux® distributions. Install libc++1-XX with XX>=14 via the Advanced Package Tool (apt).
>># apt install libc++1-19 -y
To install the package, use :
pip install <TwinCatInstallDir>\Functions\TF38xx-Machine-Learning\Utilities\ModelManagerAPI\PythonPackage\<whl-file-name>
![]() | The TwinCatInstallDir folder may contain different versions of the package (if you have installed a new setup on top of an old TwinCAT Machine Learning Setup). Make sure you always use the current version. |
Using the API
The package is loaded with
import beckhoff.toolbox as tb
Creating a JSON and PlcOpenXml from an ONNX file
tb.onnxprep("C:\\PathTo\\myONNX.onnx")
Display of model information
tb.info("C:\\PathTo\\myONNX.json")
Add Custom Attributes
new_ca = { 'nID' : -34234, 'bTested' : True, 'fNum' : 324.3E-12, 'AnotherTreeItem' : { 'fPi' : 3.13412, 'bFalseFlag' : False }}
tb.modify_ca("C:\\PathTo\\myONNX.json","C:\\PathTo\\myONNX_ca.json", new_ca)
Add Model Description (name, version, author, etc. of a model)
model_description = {
"new_version" : "2.3.1.0",
"new_name" : "CurrentPreControlAxis42",
"new_desc":"This is the most awesome model to control Axis42",
"new_author":"Max",
"new_tags":"awesome, ingenious, astounding",
}
tb.modify_md("C:\\PathTo\\myONNX.json", "C:\\PathTo\\myONNX_md.json",
**model_description)