search

CompPhysVienna / n2p2

n2p2 - A Neural Network Potential Package
https://compphysvienna.github.io/n2p2/
GNU General Public License v3.0
217 stars 82 forks source link

Pynnp #117

Closed mozhde closed 2 years ago

mozhde commented 3 years ago

The question I am planning to use the "Pynnp" library. To do so, I started with running the examples in "Pynnp/prediction"

I did a test; I ran the exact python example (the same input.nn, input.data, etc) and got the Reference and NNP energies as follow:

Energy (Ref) : -1225.22167939 Energy (NNP) : 226.6527928086986 then I did the same calculation using "nnp-predict" and got

Energy (Ref): -1.22522168E+03 Energy(NNP) : -1.22409045E+03

As you can see the prediction using these two methods are very different; and I couldn't figure it out why. I was wondering if you could help me with it?

Thanks

singraber commented 2 years ago

Hello and sorry for the late reply..

I suspect that you are using the example-verbose.py script, which was written before the data set normalization feature was added to the C++ code. Hence, a few calls need to be added to the script and in order to be able to do that I have to add stuff to the C++/Python wrapper. It will not be a lot of work, I will write a fix in the next days...

singraber commented 2 years ago

I just fixed this here: https://github.com/CompPhysVienna/n2p2/commit/6cc948bb052718b6d4225d62efb33b9d3abdef78, it should work now in the current master branch.