NNP train issues (nnp-scaling and nnp-train)

[yliu1240]

NNP train issues (nnp-scaling and nnp-train)

[YunxingZuo]

Thanks for the heads up! @yliu1240

a. The change is perfectly correct. We have also made the change to the _nnp.py and set the bin number to 100 since the different bin numbers wouldn't change the file needed to run subsequent nnp-train (scaling.data). We also found that as long as we have a second argument for nnp-scaling (e.g., input_filename), the scaling.data still showed up and didn't impact the execution of nnp-train (though with an error message).

b. The bin number seems to change the histogram counts in each separate symmetry function file. Sadly, I couldn't find the exact usage of these separate files in the tutorial as well.

c. The error output mechanism has been fixed. Please take a trial.

d. We are aware that the nnp-scaling/nnp-training supports parallel computing. We did not incorporate the 'mpirun' in the code for a more general-purpose application because not all the platforms have open-mpi installed. However, we can easily add the 'mpirun' in the code manually.

Regards