def ani_ase_calculator(ani_model):
"""
Return a calculator from a model.
choices are %s
""" % " ".join(models.keys())
return ANIENS(aniensloader(models[ani_model]))
def parse_args():
"""
parse commandline arguments
"""
parser = argparse.ArgumentParser(
description="Run ani on a testset",
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('datadir',
type=str,
help='directory holding json files to use as ref data')
parser.add_argument('-network-type',
type=str,
choices=models.keys(),
required=True)
parser.add_argument(
'-implementation',
type=str,
choices=IMPLEMENTATIONS,
default=TORCHANI,
)
parser.add_argument('-optimize',
action='store_true',
help='optimize molecule prior to computing energy')
parser.add_argument('-numb-networks',
default=8,
type=int,
help='number of committee members to load')
return parser.parse_args()
def khan_calculator(ani_model, khan_network, numb_networks):
"""
Return a calculator from a roitberg model.
choices are %s
""" % " ".join(models.keys())
if khan_network is None:
print("setting up Roitberg network")
model_file = models[ani_model]
wkdir = model_file.rsplit('/', 1)[0] + '/'
data = np.loadtxt(model_file, dtype=str)
cnstfile = wkdir + data[0] # AEV parameters
saefile = wkdir + data[1] # Atomic shifts
nnfdir = wkdir + data[2] # network prefix
Nn = int(data[3]) # Number of networks in the ensemble
assert numb_networks <= Nn
network_dir = nnfdir[:-5]
else:
print("setting up khan network")
saefile = os.path.join(os.environ["KHAN"], "data", "sae_linfit.dat"
#"sae_wb97x.dat"
)
network_dir = None
ani_lib = os.path.join(os.environ["KHAN"], "gpu_featurizer", "ani_cpu.so")
initialize_module(ani_lib)
cp = tf.ConfigProto(
intra_op_parallelism_threads=1,
inter_op_parallelism_threads=1,
)
tf_sess = tf.Session(config=cp)
atomic_energies = read_sae_parameters(saefile)
calculator = Calculator(tf_sess,
atomic_energies,
numb_networks,
khan_saved_network=khan_network,
roitberg_network=network_dir)
return calculator
def parse_args():
"""
parse commandline arguments
"""
parser = argparse.ArgumentParser(
description="optimize structures with ANI",
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument(
'mae_file',
type=str,
help='name of mae file',
)
parser.add_argument(
'-network-type',
type=str,
choices=models.keys(),
default="ani-1x",
)
parser.add_argument(
'-implementation',
type=str,
choices=IMPLEMENTATIONS,
default=KHAN,
)
parser.add_argument('-khan-network',
type=str,
default=None,
help='khan trained network')
parser.add_argument('-optimize', action='store_true', help='optimize')
parser.add_argument('-numb-networks',
default=8,
type=int,
help='number of committee members to load')
return parser.parse_args()
def parse_args():
"""
parse commandline arguments
"""
parser = argparse.ArgumentParser(
description="optimize a crystal structure from a cif file with ANI",
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('cif_file',
type=str,
help='name of cif file (must end in .cif)')
parser.add_argument('-network-type',
type=str,
choices=models.keys(),
required=True)
parser.add_argument(
'-implementation',
type=str,
choices=IMPLEMENTATIONS,
default=TORCHANI,
)
parser.add_argument('-khan-network',
type=str,
default=None,
help='khan trained network')
parser.add_argument('-numb-networks',
default=8,
type=int,
help='number of committee members to load')
parser.add_argument('-MC',
type=str,
required=False,
help='running Monte Carlo for n steps')
return parser.parse_args()
def torchani_calculator(ani_model, numb_networks):
"""
Return a calculator from a model.
choices are %s
""" % " ".join(models.keys())
model_file = models[ani_model]
wkdir = model_file.rsplit('/', 1)[0] + '/'
data = np.loadtxt(model_file, dtype=str)
cnstfile = wkdir + data[0] # AEV parameters
saefile = wkdir + data[1] # Atomic shifts
nnfdir = wkdir + data[2] # network prefix
Nn = int(data[3]) # Number of networks in the ensemble
assert numb_networks <= Nn
constants = torchani.neurochem.Constants(cnstfile)
energy_shifter = torchani.neurochem.load_sae(saefile)
aev_computer = torchani.AEVComputer(**constants)
aev_computer.to(torch.double)
nn_models = torchani.neurochem.load_model_ensemble(
constants.species,
nnfdir,
numb_networks)
# go to double precision
for model in nn_models:
model.to(torch.double)
calculator = torchani.ase.Calculator(
constants.species,
aev_computer,
nn_models,
energy_shifter)
return calculator
def parse_args():
"""
parse commandline arguments
"""
parser = argparse.ArgumentParser(
description="optimize a crystal structure from a cif file with ANI",
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('cif_file',
type=str,
help='name of cif file (must end in .cif)')
parser.add_argument('-network-type',
type=str,
choices=models.keys(),
required=True)
parser.add_argument(
'-implementation',
type=str,
choices=IMPLEMENTATIONS,
default=TORCHANI,
)
parser.add_argument('-khan-network',
type=str,
default=None,
help='khan trained network')
parser.add_argument('-numb-networks',
default=8,
type=int,
help='number of committee members to load')
parser.add_argument('-MC',
type=int,
required=False,
help='running Monte Carlo for n steps')
parser.add_argument(
'-space_group',
default=1,
type=int,
help=
'under which space group you would like to run, by default it is P1')
parser.add_argument(
'-step_size',
dest='MC_f1_f2_f3_f4_f5_f6_f7_f8',
metavar='<Xi>',
default=(0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1),
nargs=8,
type=float,
help=
'MC step size for translate, rotatioin, cell length(a,b,c), and cell angle(alpha, beta, gamma), unit(A or degree)'
)
parser.add_argument(
'-perturbation_type',
choices=[
'translation', 'rotation', 'cell_length_a', 'cell_length_b',
'cell_length_c', 'cell_angle_alpha', 'cell_angle_beta',
'cell_angle_gamma'
],
type=str,
default=[
'translation', 'rotation', 'cell_length_a', 'cell_length_b',
'cell_length_c', 'cell_angle_alpha', 'cell_angle_beta',
'cell_angle_gamma'
],
help=
'choose one type of perturbation, by default run all types randomly')
parser.add_argument(
'-debug',
default=False,
dest='debug',
action='store_true',
help='debug option to print out all the accepted frames')
return parser.parse_args()
def parse_args():
"""
parse commandline arguments
"""
parser = argparse.ArgumentParser(
description="optimize a crystal structure from a cif file with ANI",
formatter_class=argparse.ArgumentDefaultsHelpFormatter
)
parser.add_argument(
'cif_file',
type=str,
help='name of cif file (must end in .cif)'
)
parser.add_argument(
'-network-type',
type=str,
choices=models.keys(),
required=True
)
parser.add_argument(
'-implementation',
type=str,
choices=IMPLEMENTATIONS,
default=TORCHANI,
)
parser.add_argument(
'-khan-network',
type=str,
default=None,
help='khan trained network'
)
parser.add_argument(
'-numb-networks',
default=8,
type=int,
help='number of committee members to load'
)
parser.add_argument(
'-MC',
type=int,
required=False,
help='running Monte Carlo for n steps'
)
parser.add_argument(
'-space_group',
default=1,
type=int,
help='under which space group you would like to run, by default it is P1'
)
parser.add_argument(
'-step_size',
dest='MC_f1_f2_f3_f4',
metavar='<Xi>',
default=(0.1, 0.1, 0.1, 0.1),
nargs=4,
type=float,
help='MC step size for translate, rotatioin, cell length, and cell angle, unit(A or degree)'
)
return parser.parse_args()
def parse_args():
"""
parse commandline arguments
"""
parser = argparse.ArgumentParser(
description="optimize a crystal structure from a cif file with ANI",
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('cif_file',
type=str,
help='name of cif file (must end in .cif)')
parser.add_argument('-network-type',
type=str,
choices=models.keys(),
required=True)
parser.add_argument('-time-step',
default=0.1,
type=float,
help="time step in fs")
parser.add_argument('-total-time',
default=10.0,
type=float,
help="total MD time in ps")
parser.add_argument(
'-implementation',
type=str,
choices=IMPLEMENTATIONS,
default=TORCHANI,
)
parser.add_argument('-khan-network',
type=str,
default=None,
help='khan trained network')
parser.add_argument('-optimize',
action='store_true',
help='optimize crystal prior to computing energy')
parser.add_argument(
'-relax-cell',
action='store_true',
help='allow cell shape to change when optimizing',
)
parser.add_argument('-dynamics',
action='store_true',
help='perform molecular dynamics')
parser.add_argument('-numb-networks',
default=8,
type=int,
help='number of committee members to load')
parser.add_argument('-truncate-box',
default=False,
action="store_true",
help='truncate a box at 12 angstrom')
parser.add_argument(
'-torsion',
default=None,
nargs=5,
help=
"list of one float and four integers defining a torsion to scan from a value of zero"
)
parser.add_argument(
'-box-length',
type=float,
default=None,
help='if system is not periodic, place it in a box of this length')
parser.add_argument(
'-supercell',
type=int,
default=None,
help='make a supercell of this size from initial system')
return parser.parse_args()