atomistic-machine-learning/schnetpack

How to use other properties to predict the energy and forces?

Closed this issue · 1 comments

How can I use the following data as input of the model please?

elements: A list with length equal to the number of atoms. Each of these elements indicates the atom type. For example, elements=[H,H,0'] for a water molecule.

coordinates: A list with length equal to the number of atoms. Among them, the i-th is a 3-tuple representing the three-dimensional coordinates (x, y, z) of the i-th atom.

connectivity: A list with length equal to the number of atoms. Where the i-th element is a list of all connected atoms of the i-th atom.

bond type: A list with length equal to the number of keys. The i-th element represents the type of the i-th key.

formal charge: A list with length equal to the number of atoms. The i-th element represents the formal charge of the i-th atom, represented by a floating-point number, a floating-point number. molecular energy to predict

If I cannot put all of them into the model, which can I take use? Thanks!

Hi @xiehuanyi
schnetpack generally uses atom types and coordinates for learning physical properties.
So without any further changes to the code you can use elements and coordinates.
Best,
Stefaan