GaPFlow.md.MolecularDynamics

class GaPFlow.md.MolecularDynamics

Bases: object

Driver for molecular dynamics simulations.

Abstract base class for MD setup, running, and reading outputs. Derived classes need to implement methods to write LAMMPS input files into a dtool dataset, and to read the output of this simulation.

name

Name of the MD object

Type:

str

params

Parameters to control the MD setup, will be written to the dtool metadata.

Type:

dict

main_file

File name of the main LAMMPS input file.

Type:

str

num_worker

Number of cores to run the parallel MD simulation.

Type:

int

is_mock

Whether the subclass is only a mock object, which does not run an actual MD simulation.

Type:

bool

__init__()

Methods

__init__()
build_input_files(dataset, location, X)	Builds LAMMPS input files based on GP inputs
read_output()	Read simulation output and returns observations and their standard error.
run(X_target, tag)	Run an MD simulation and store its input, metadata, and output into a dtool dataset.

Attributes

dtool_basepath	File location, where dtool datasets are written into (default is '/tmp/').
params
main_file
num_worker
is_mock

abstractmethod build_input_files(dataset, location, X) → None

Builds LAMMPS input files based on GP inputs

and writes them to a dtool dataset.

Parameters:

• dataset (dtoolcore.proto_dataset) – A proto_dataset object.

• location (str) – Absolute path of the proto dataset.

• X (Array) – Input (i.e. density, gap height, …)

property dtool_basepath

File location, where dtool datasets are written into (default is ‘/tmp/’).

is_mock: bool

main_file: str

name

alias of str

num_worker: int

params: dict

abstractmethod read_output()

Read simulation output and returns observations and their standard error.

run(X_target, tag)

Run an MD simulation and store its input, metadata, and output into a dtool dataset.

This method is called from a Database instance when new training data is added e.g. during initialization or in an active learning simulation.

Parameters:

• X_target (Array) – The training input.

• tag (str) – A tag to attach to the dataset name.

Returns:

• Array – Training observations

• Array – Standard error of training observations