It doesn't use the pip module but old system: NumbaWarning: ..requires TBB version

[Pyan]

One Python program I want to use drops the warning:

NumbaWarning: The TBB threading layer requires TBB version 2021 update 6 or later i.e., TBB_INTERFACE_VERSION >= 12060. Found TBB_INTERFACE_VERSION = 12050. The TBB threading layer is disabled

I did installed the tbb module 2021.11.0 with pip install (pip show tbb).
However my Linux (Mint 21.2) system has an older version of it:

pkg-config --modversion tbb
2021.5.0

It looks like the running Python program uses the system module and not the module installed with pip.
I also tried pipenv (shell). It did not work either.

I tried to remove the system libtbb but after this an error drops:

ImportError: libopencv_imgproc.so.4.5d: cannot open shared object file: No such file or directory

I can fix it with installing 'libopencv-dev' at the system again. And with this it also installs tbb 2021.5.0 to the system.
Just doing pip install opencv-python did not worked out.

I can do

export NUMBA_THREADING_LAYER=omp

After that no warning appears. But I'm not sure how different the performance is in between omp and tpp. Using omp instead of none has about not impact to performance (both about 2h run time).

If I do

export NUMBA_THREADING_LAYER=tpp

I get the error:

ValueError: No threading layer could be loaded.
HINT:
Intel TBB is required, try:
$ conda/pip install tbb

But I already did 'pip install tbb'.

If I use pipenv the folder for the environment is listed last in sys.paths. Manually changing it to first position did not work either.

Any other ideas to resolve this?
(I would like to avoid installing packages to system from external sources but the Linux mint package manager.)