Possibly tensorflow related installation issue

[dominiquesydow]

I also noticed that I couldn't install package teachopencadd where was function seed_everything () because of the tensorflow-base.2.18.0 (something like this). Conda told me that I had corrupted file, pip told me that there was parse-function problem(if I remember correctly). I only want to say that I could resolve this problem with the help of pip install pytorch-lightning=1.7.7, where pytorch-lightning utilities.seed had the same function.

Originally posted by @Q84-wq in #419 (comment)

[dominiquesydow]

Hi @AndreaVolkamer,

Can anyone in your lab please take a look at this?
If tensorflow is a source of problems again, I wonder if we could remove T021 and T022 from our global teachopencadd package environment.

Why?
We have had so many issues already with having tensorflow as part of the installation and afaiu tensorflow is not really the go-to ML/DL package anymore. So why not instead add an installation only for these two talktorials (maybe really with pinned versions).

[dominiquesydow]

Hi @Q84-wq,

Could you please add some information on this issue: (a) your OS and (b) how you tried to install teachopencadd?

I just tried on MacOS via
CONDA_SUBDIR=osx-64 mamba create -n teachopencadd teachopencadd
and was not able to reproduce the error you described. But that might just be because your setup was different.

Many thanks!

[Q84-wq]

Hello, miss Sydow,
I run my laptop on Windows 64-bit (Lenovo) and the app is Jupyter Notebook desktop. I tried again to install teachopencadd. Firstly I installed mamba into conda (I understand that it's not recommended, but it's more convenient for me) https://mamba.readthedocs.io/en/latest/installation/mamba-installation.html, then I used
mamba create -n teachopencadd teachopencadd -c conda-forge -c defaults
and there was written "warning libmamba Problem type not implement SOLVER_RULE_STRICT_REPO_PRIORITY" many times,
and then "Could not solve the environment specs
The following packages are incompatible" and there were biopython - > python, numpy; biotite - > python, h5py 2.10.0 - > python_abi 3.6(3.7).* *_cp36(37)m, mdanalysis - > python_abi; tensorflow 1.14.0, numpy - > python
The second way, mamba env create -f https://raw.githubusercontent.com/volkamerlab/TeachOpenCADD/master/devtools/test_env.yml
conda activate teachopencadd, it has stuck on Solving environment for a really long time
Thank you very much!