I followed the advices in the troubleshooting from
I did so because I needed to install pandas in a version that requires numpy to be more recent than the one that comes with FME.
So far I thought all went fine, but when I call numpy.__version__ from within a PythonCreator then the version number returned is 1.26.2. When I use pip.exe list from within my custom Python installation folder then the numpy version returned is 2.2.6 → matching the version number from when I installed numpy through pip.
I also installed pandas and SQLAlchemy through pip and the version numbers returned from the PythonCreator do match the version numbers pip.exe list returns.
I wonder why numpy seems not to be used from the custom Python installation but rather from a fall back version that comes with FME Flow. Did I forget some step?
- I installed Python 3.12 on the FME Flow machine
- I ran the following three commands with the paths:
-
<FMEFlowDir>Server/fme/fme.exe APPLY_SETTINGS SYSTEM "Python/Use Custom Python 64" true
-
<FMEFlowDir>Server/fme/fme.exe APPLY_SETTINGS SYSTEM "Python/Python Interpreter 64" c:/<path_to_dll>.dll
-
<FMEFlowDir>Server/fme/fme.exe APPLY_SETTINGS SYSTEM "Python/Custom Python Home 64" c:/<path_to_python_home>
-
-
My Python Compatibility setting in the published workspace is set to ‘Python 3.12+’
I added the workspace with the PythonCreator, just in case.