fortran modules checked in model?

[andrew_peterson]

Why are the fortran modules checked for the correct version in model/__init__.py? It seems that they are used in other areas of the code: particularly descriptors. Shouldn't this be done in the __init__.py file of the amp module?

Note it's possible to use just the descriptor part of Amp without a model to generate fingerprints, etc. (These can then be fed into another machine learning model or just used for analysis.)

[Alireza Khorshidi]

Currently, we do not use fmodules in __init__.py. We just use it in model/__init__.py as well as descriptor/gaussian.py and descriptor/zernike.py. Where do you suggest we check for fmodules version?

[andrew_peterson]

Well, if some one just uses descriptor/gaussian.py to make fingerprints, that they then use in another learning model (like Zack did in tensorflow), their code would never check to see if fmodules is updated. So either it should be in every import line or else checked once in init.py?

[Alireza Khorshidi]

I agree. Moved to amp/init.py in the commit 5893bdd.