pinned to TF2.4

[alisterburt]

Hi, I was wondering how strict the pin to TF==2.4 is? I would like to use cryoCARE but don't want to restrict users to 3.8 and TF==2.4 is not available on newer versions of Python

[thorstenwagner]

I had quite some trouble to get it running with TF 2.4 and needed a lot of try and error... :-/

[schloegl]

For other versions of tensorflow, the compatibility between tensorflow/cuda/cudnn need to be be considered:
https://www.tensorflow.org/install/source#gpu

For newer versions of numpy,
"numpy.bool" need to be replaced with "numpy.bool_"
https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations

I'm currently installing with
python/3.10, tensorflow 2.14.1, cuda/11.8, cudnn/8.7, numpy/1.26.4
and
python/3.11, tensorflow 2.15 , cuda/12.2, cudnn/8.8, numpy/1.23
after applying these changes:
master...schloegl:cryoCARE_pip:master