"""ASE Calculator for the ground state exciting DFT code.
Exciting calculator class in this file allow for writing exciting input
files using ASE Atoms object that allow for the compiled exciting binary
to run DFT on the geometry/material defined in the Atoms object. Also gives
access to developer to a lightweight parser (lighter weight than NOMAD or
the exciting parser in the exciting repository) to capture ground state
properties.
Note: excitingtools must be installed using `pip install excitingtools` to
use this calculator.
"""
from os import PathLike
from pathlib import Path
from typing import Any, Mapping
import ase.io.exciting
from ase.calculators.calculator import PropertyNotImplementedError
from ase.calculators.exciting.runner import (SimpleBinaryRunner,
SubprocessRunResults)
from ase.calculators.genericfileio import (BaseProfile, CalculatorTemplate,
GenericFileIOCalculator)
class ExcitingProfile(BaseProfile):
"""Defines all quantities that are configurable for a given machine.
Follows the generic pattern BUT currently not used by our calculator as:
* species_path is part of the input file in exciting.
* OnlyTypo fix part of the profile used in the base class is the run
method, which is part of the BinaryRunner class.
"""
def __init__(self, binary, species_path=None, **kwargs):
super().__init__(**kwargs)
self.species_path = species_path
self.binary = binary
def version(self):
"""Return exciting version."""
# TARP No way to get the version for the binary in use
return
# Machine specific config files in the config
# species_file goes in the config
# binary file in the config.
# options for that, parallel info dictionary.
# Number of threads and stuff like that.
def get_calculator_command(self, input_file):
"""Returns command to run binary as a list of strings."""
# input_file unused for exciting, it looks for input.xml in run
# directory.
if input_file is None:
return [self.binary]
else:
return [self.binary, str(input_file)]
class ExcitingGroundStateTemplate(CalculatorTemplate):
"""Template for Ground State Exciting Calculator
Abstract methods inherited from the base class:
* write_input
* execute
* read_results
"""
parser = {'info.xml': ase.io.exciting.parse_output}
output_names = list(parser)
# Use frozenset since the CalculatorTemplate enforces it.
implemented_properties = frozenset(['energy', 'forces'])
_label = 'exciting'
def __init__(self):
"""Initialise with constant class attributes.
:param program_name: The DFT program, should always be exciting.
:param implemented_properties: What properties should exciting
calculate/read from output.
"""
super().__init__('exciting', self.implemented_properties)
self.errorname = f'{self._label}.err'
@staticmethod
def _require_forces(input_parameters):
"""Expect ASE always wants forces, enforce setting in input_parameters.
:param input_parameters: exciting ground state input parameters, either
as a dictionary or ExcitingGroundStateInput.
:return: Ground state input parameters, with "compute
forces" set to true.
"""
from excitingtools import ExcitingGroundStateInput
input_parameters = ExcitingGroundStateInput(input_parameters)
input_parameters.tforce = True
return input_parameters
def write_input(
self,
profile: ExcitingProfile, # ase test linter enforces method signatures
# be consistent with the
# abstract method that it implements
directory: PathLike,
atoms: ase.Atoms,
parameters: dict,
properties=None,
):
"""Write an exciting input.xml file based on the input args.
:param profile: an Exciting code profile
:param directory: Directory in which to run calculator.
:param atoms: ASE atoms object.
:param parameters: exciting ground state input parameters, in a
dictionary. Expect species_path, title and ground_state data,
either in an object or as dict.
:param properties: Base method's API expects the physical properties
expected from a ground state calculation, for example energies
and forces. For us this is not used.
"""
# Create a copy of the parameters dictionary so we don't
# modify the callers dictionary.
parameters_dict = parameters
assert set(parameters_dict.keys()) == {
'title', 'species_path', 'ground_state_input',
'properties_input'}, \
'Keys should be defined by ExcitingGroundState calculator'
file_name = Path(directory) / 'input.xml'
species_path = parameters_dict.pop('species_path')
title = parameters_dict.pop('title')
# We can also pass additional parameters which are actually called
# properties in the exciting input xml. We don't use this term
# since ASE use properties to refer to results of a calculation
# (e.g. force, energy).
if 'properties_input' not in parameters_dict:
parameters_dict['properties_input'] = None
ase.io.exciting.write_input_xml_file(
file_name=file_name, atoms=atoms,
ground_state_input=parameters_dict['ground_state_input'],
species_path=species_path, title=title,
properties_input=parameters_dict['properties_input'])
def execute(
self, directory: PathLike,
profile) -> SubprocessRunResults:
"""Given an exciting calculation profile, execute the calculation.
:param directory: Directory in which to execute the calculator
exciting_calculation: Base method `execute` expects a profile,
however it is simply used to execute the program, therefore we
just pass a SimpleBinaryRunner.
:param profile: This name comes from the superclass CalculatorTemplate.
It contains machine specific information to run the
calculation.
:return: Results of the subprocess.run command.
"""
return profile.run(directory, f"{directory}/input.xml", None,
erorrfile=self.errorname)
def read_results(self, directory: PathLike) -> Mapping[str, Any]:
"""Parse results from each ground state output file.
Note we allow for the ability for there to be multiple output files.
:param directory: Directory path to output file from exciting
simulation.
:return: Dictionary containing important output properties.
"""
results = {}
for file_name in self.output_names:
full_file_path = Path(directory) / file_name
result: dict = self.parser[file_name](full_file_path)
results.update(result)
return results
def load_profile(self, cfg, **kwargs):
"""ExcitingProfile can be created via a config file.
Alternative to this method the profile can be created with it's
init method. This method allows for more settings to be passed.
"""
return ExcitingProfile.from_config(cfg, self.name, **kwargs)
class ExcitingGroundStateResults:
"""Exciting Ground State Results."""
def __init__(self, results: dict) -> None:
self.results = results
self.final_scl_iteration = list(results['scl'].keys())[-1]
def total_energy(self) -> float:
"""Return total energy of system."""
# TODO(Alex) We should a common list of keys somewhere
# such that parser -> results -> getters are consistent
return float(
self.results['scl'][self.final_scl_iteration]['Total energy']
)
def band_gap(self) -> float:
"""Return the estimated fundamental gap from the exciting sim."""
return float(
self.results['scl'][self.final_scl_iteration][
'Estimated fundamental gap'
]
)
def forces(self):
"""Return forces present on the system.
Currently, not all exciting simulations return forces. We leave this
definition for future revisions.
"""
raise PropertyNotImplementedError
def stress(self):
"""Get the stress on the system.
Right now exciting does not yet calculate the stress on the system so
this won't work for the time being.
"""
raise PropertyNotImplementedError
[docs]class ExcitingGroundStateCalculator(GenericFileIOCalculator):
"""Class for the ground state calculation.
:param runner: Binary runner that will execute an exciting calculation and
return a result.
:param ground_state_input: dictionary of ground state settings for example
{'rgkmax': 8.0, 'autormt': True} or an object of type
ExcitingGroundStateInput.
:param directory: Directory in which to run the job.
:param species_path: Path to the location of exciting's species files.
:param title: job name written to input.xml
:return: Results returned from running the calculate method.
Typical usage:
gs_calculator = ExcitingGroundState(runner, ground_state_input)
results: ExcitingGroundStateResults = gs_calculator.calculate(
atoms: Atoms)
"""
def __init__(
self,
*,
runner: SimpleBinaryRunner,
ground_state_input,
directory='./',
species_path='./',
title='ASE-generated input',
parallel=None,
parallel_info=None,
):
self.runner = runner
# Package data to be passed to
# ExcitingGroundStateTemplate.write_input(..., input_parameters, ...)
# Structure not included, as it's passed when one calls .calculate
# method directly
self.exciting_inputs = {
'title': title,
'species_path': species_path,
'ground_state_input': ground_state_input,
}
self.directory = Path(directory)
# GenericFileIOCalculator expects a `profile`
# containing machine-specific settings, however, in exciting's case,
# the species file are defined in the input XML (hence passed in the
# parameters argument) and the only other machine-specific setting is
# the BinaryRunner. Furthermore, in GenericFileIOCalculator.calculate,
# profile is only used to provide a run method. We therefore pass the
# BinaryRunner in the place of a profile.
super().__init__(
profile=runner,
template=ExcitingGroundStateTemplate(),
directory=directory,
parameters=self.exciting_inputs,
parallel_info=parallel_info,
parallel=parallel,
)