def main(args):
mol_list, output_list = readTestData(args)
if os.path.exists("./cod_predict.db"):
os.remove("./cod_predict.db")
new_dataset = AtomsData('./cod_predict.db',
available_properties=['band_gap'])
print('Number of test instances ' + str(len(output_list)))
new_dataset.add_systems(mol_list, output_list)
def convert_to_db(file_name):
mol_list = []
property_list = []
atoms = read(os.path.join("./", file_name), index=':')
property_list.append({'band_gap': np.array([-97208.40600498248], dtype=np.float32)})
mol_list.extend(atoms)
if os.path.exists("./cod_predict.db"):
os.remove("./cod_predict.db")
new_dataset = AtomsData('./cod_predict.db', available_properties=['band_gap'])
new_dataset.add_systems(mol_list, property_list)
def _write_database(self):
if len(self.samples) > 0:
dataset = AtomsData(
self.dataset, available_properties=self.samples_thresholds[0]
)
dataset.add_systems(self.samples, self.samples_thresholds)
logging.info(
"{:d} samples written to {:s}.".format(len(self.samples), self.dataset)
)
else:
logging.info("No samples collected.")
def get_all_features_spatial(df, end):
df = shuffle(df)
df = df.reset_index(drop=True)
df = df[:end]
xyz_all = ''
for i, row in df.iterrows():
#print(row['h**o'])
xyz = row['xyz']
xyz_new = xyz.split("\n", 2)[0] + '\n' + str(
row['h**o']) + '\n' + xyz.split("\n", 2)[2]
xyz_all = xyz_all + xyz_new
with open("coord.xyz", "w") as xyz_file:
xyz_file.write(xyz_all)
atoms = read('coord.xyz', index=':10')
property_list = []
for at in atoms:
# All properties need to be stored as numpy arrays.
# Note: The shape for scalars should be (1,), not ()
# Note: GPUs work best with float32 data
h**o = np.array([float(list(at.info.keys())[0])], dtype=np.float32)
property_list.append({'h**o': h**o})
#print('Properties:', property_list)
new_dataset = AtomsData('./new_dataset.db', available_properties=['h**o'])
new_dataset.add_systems(atoms, property_list)
'''
print('Number of reference calculations:', len(new_dataset))
print('Available properties:')
for p in new_dataset.available_properties:
print('-', p)
print()
example = new_dataset[0]
print('Properties of molecule with id 0:')
for k, v in example.items():
print('-', k, ':', v.shape)
'''
return (new_dataset)
def gnn_pred(cif_file):
# device = torch.device("cuda" if args.cuda else "cpu")
device = "cpu"
sch_model = torch.load(os.path.join("./schnetpack/model", 'best_model'),
map_location=torch.device(device))
test_dataset = AtomsData('./cod_predict.db')
test_loader = spk.AtomsLoader(test_dataset, batch_size=32)
prediction_list = []
for count, batch in enumerate(test_loader):
# move batch to GPU, if necessary
print('before batch')
batch = {k: v.to(device) for k, v in batch.items()}
print('after batch')
# apply model
pred = sch_model(batch)
prediction_list.extend(
pred['band_gap'].detach().cpu().numpy().flatten().tolist())
return prediction_list[0]
args = arg_parser.parse_args()
run_params = args.__dict__
params_hash = hashlib.sha256(
json.dumps(run_params).encode()).hexdigest()[:6]
# Determine the output directory
test_dir = os.path.join(
'networks',
f'T{args.num_messages}_b{args.batch_size}_n{args.num_epochs}_{params_hash}'
)
os.makedirs(test_dir)
with open(os.path.join(test_dir, 'config.json'), 'w') as fp:
json.dump(run_params, fp)
# Making the data loaders
train_data = AtomsData('datasets/train.db')
train_loader = AtomsLoader(train_data,
args.batch_size,
shuffle=True,
pin_memory=True,
num_workers=2)
test_data = AtomsData('datasets/test.db')
test_loader = AtomsLoader(test_data, args.batch_size)
valid_data = AtomsData('datasets/valid.db')
valid_loader = AtomsLoader(valid_data,
args.batch_size,
pin_memory=True,
num_workers=2)
# Make the model
mean, std = train_loader.get_statistics('ip',
def load_dataset(self):
''' Loads the dataset and stores it in `.dataset`
Currently supported:
xyz format - needs to be extended with energy in the comment and forces
npz format - as given by sGDML, needs to contain 'R', 'E', 'F'
db format - as given by schnetpack
'''
path = self.args['dataset_file']
if path is None:
print_error(
f"No dataset given. Please use the -d arg followed by the path to the dataset."
)
elif not os.path.exists(path):
print_error(f"Dataset path {path} is not valid.")
ext = os.path.splitext(path)[-1]
#xyz file
if ext == ".xyz":
print_ongoing_process(f"Loading xyz file {path}")
try:
file = open(path)
dat = read_concat_ext_xyz(file)
data = {
'R': np.array(dat[0]),
'z': dat[1],
'E': np.reshape(dat[2], (len(dat[2]), 1)),
'F': np.array(dat[3])
}
except Exception as e:
print(e)
print_error("Couldn't load .xyz file.")
print_ongoing_process(f"Loaded xyz file {path}", True)
#npz file
elif ext == ".npz":
print_ongoing_process(f"Loading npz file {path}")
try:
data = np.load(path, allow_pickle=True)
except Exception as e:
print(e)
print_error("Couldn't load .npz file.")
print_ongoing_process(f"Loaded npz file {path}", True)
# schnetpack .db
elif ext == '.db':
print_ongoing_process(f"Loading db file {path}")
from schnetpack import AtomsData
data = AtomsData(path)
print_ongoing_process(f"Loaded db file {path}", True)
else:
print_error(
f"Unsupported data type {ext} for given dataset {path} (xyz, npz, schnetpack .db supported)."
)
self.dataset = data
self.dataset_path = path
if self.get_para('load_dataset', 'post_processing') is not None:
print_ongoing_process('Post-processing dataset')
self.call_para('load_dataset', 'post_processing', args=[self])
print_ongoing_process('Post-processing dataset', True)
for at in tqdm(atoms):
# All properties need to be stored as numpy arrays.
# Note: The shape for scalars should be (1,), not ()
# Note: GPUs work best with float32 data
# print(at.info.keys())
# print(list(at.info.keys()))
# energy = np.array([float(list(at.info.keys())[0])], dtype=np.float32)
energy = np.array([float(at.info['energy'])], dtype=np.float32)
forces = np.array(at.get_forces(), dtype=np.float32)
property_list.append({'energy': energy, 'forces': forces})
# print(energy)
# print(type(energy))
# print('Properties:', property_list)
new_dataset = AtomsData('./40-cspbbr3-300K.db',
available_properties=['energy', 'forces'])
new_dataset.add_systems(atoms, property_list)
print('Number of reference calculations:', len(new_dataset))
print('Available properties:')
for p in new_dataset.available_properties:
print('-', p)
print()
example = new_dataset[0]
print('Properties of molecule with id 0:')
for k, v in example.items():
print('-', k, ':', v.shape)