def _parse_args():
parser = argparse.ArgumentParser()
parser.add_argument('model')
parser.add_argument('--model_path')
parser.add_argument('-o', '--output-dir', required=True)
parser.add_argument('-p', '--protein', default='5ht1b')
parser.add_argument('-d', '--debug', action='store_true')
parser.add_argument('-n', '--n-molecules', type=int, default=250)
parser.add_argument('-m', '--mode', default='minimize')
parser.add_argument('-r', '--random-samples', type=int, default=100)
parser.add_argument('-f', '--fine-tune-epochs', type=int, default=5)
parser.add_argument('--dataset', default='default')
parser.add_argument('--n-cpu', default=4, type=int)
args = parser.parse_args()
setup_and_get_logger(args.debug)
scripting.set_keras_cores(args.n_cpu)
disable_rdkit_logging()
if args.model not in ALL_MODELS:
logger.error(f'No model named {args.model}')
raise ValueError(f'No model named {args.model}')
if args.model_path is None:
if 'pretrained' not in ALL_MODELS[args.model]:
raise ValueError(f'No pretrained {args.model} model delivered. '
'Provide the path to pretrained model.')
args.model_path = ALL_MODELS[args.model]['pretrained']
args.output_dir = os.path.realpath(args.output_dir)
if args.debug:
args.n_molecules = 2
args.fine_tune_epochs = 0
args.random_samples = 2
logger.debug(
'Arguments updated due to debug mode on, current args: %s',
str(args))
return args
'mse', # loss function to use with property prediction error for regression tasks
'logit_prop_pred_loss': 'binary_crossentropy',
# loss function to use with property prediction for logistic tasks
'prop_pred_loss_weight': 0.5,
'prop_pred_dropout': 0.0,
'prop_batchnorm': True,
# print output parameters
"verbose_print": 0,
'MAX_LEN': 200,
'RAND_SEED': 0,
'PADDING': 'right'
}
logger = setup_and_get_logger(name=__name__)
def _log_loaded_params(params):
logger.info('CVAE overwritten hyper-parameters:')
for key, value in params.items():
logger.info('{:25s} - {:12}'.format(key, str(value)))
def load_params(param_file=None, verbose=True):
if param_file is None:
return deepcopy(DEFAULT_PARAMETERS)
loaded_parameters = json.loads(open(param_file).read(),
object_pairs_hook=OrderedDict)
import argparse
import os
from datetime import datetime
from docking_baselines.datasets.loaders import Dataset
from docking_baselines.models.models import ALL_MODELS
from docking_baselines.utils.scripting import set_keras_cores
from docking_benchmark.data.directories import PRETRAINED_MODELS
from docking_benchmark.utils.logging import setup_and_get_logger
logger = setup_and_get_logger(True, __name__)
def _parse_args():
parser = argparse.ArgumentParser()
parser.add_argument('dataset')
parser.add_argument('-b', '--batch-size', type=int, default=64)
parser.add_argument('-d', '--debug', action='store_true')
parser.add_argument('-e', '--epochs', type=int, default=50)
parser.add_argument('-m', '--mode', default='generator')
parser.add_argument('--n-cpu', type=int, default=4)
parser.add_argument('-s', '--save-path')
arguments = parser.parse_args()
set_keras_cores(arguments.n_cpu)
if arguments.save_path is None:
arguments.save_path = os.path.join(
PRETRAINED_MODELS,
'gvae-' + arguments.dataset + '-' + datetime.today().strftime('%Y-%m-%d-%H:%M:%S')
)