Python environment on NERSC for IPS+OMFIT

[parkjm]

@elwasif @orso82 @smithsp

omfit on cori uses a local miniconda installation:

/global/cfs/cdirs/atom/atom_repo/miniconda3_2.1.0_11/bin/python

The default python environment of IPS is

module load python (==> python/3.7-anaconda-2019.10)

In order to use omfit classes in the ips wrrapper, 1) we have to use @orso82 ’s miniconda install or 2) install omfit dependancies under conda or virtualenv for a NERSC managed python.

Suggestion?

[orso82]

If you are only going to use certain classes, you may get away with a reduced set of OMFIT dependencies. That said, it may just be easier to either use the OMFIT's miniconda installation, or install OMFIT's required https://github.com/gafusion/OMFIT-source/blob/unstable/install/requirements.txt and optional https://github.com/gafusion/OMFIT-source/blob/unstable/install/optional.txt packages.

[elwasif]

For the IPS framework itself, it doesn't really matter which one is used. For the various workflows, it depends on which Python modules are used. I suspect the miniconda installation probably has everything most or all workflows need (the one major system specific module at NERSC that may require special tuning is mpi4py which is linked against the system MPI libraries for the default Python environments)