xnn.common.data.runner_io#
Read and write the RuNNer input.data structure format.
input.data is the data set format of the RuNNer code (Behler group) for
high-dimensional neural network potentials: a sequence of structures, each
between a begin and an end line, with lattice lines for a periodic
cell, one atom line per atom, and the energy and total charge of the
structure. RuNNer works in atomic units (bohr, hartree, elementary charges);
the reader converts to the xnn units (Angstrom, eV) and returns the structure
dicts of structure_to_graph().
The columns of the atom lines follow the begin line. A bare begin
means the default layout:
atom x y z element charge energy fx fy fz
and a begin line may name its own, with the width of a vector property in
parentheses, e.g. begin position(3) element charges hirshfeld_volume
forces(3). The reader understands position, element, charge
(charges), forces (force) and skips every other property.
Example:
from xnn.common.data.runner_io import read_runner_data
structures = read_runner_data("input.data") # Angstrom, eV, e
Module Attributes
the |
Functions
|
Iterate over the structures of a RuNNer |
|
Read every structure of a RuNNer |
|
Write structure dicts to a RuNNer |