def main():
input_folder = Path('data/input/')
train_file = Path('train.csv')
test_file = Path('test.csv')
logging.info('> Loading data')
train = pd.read_csv(input_folder/train_file, index_col=0)
test = pd.read_csv(input_folder/test_file, index_col=0)
train = reduce_mem_usage(train)
test = reduce_mem_usage(test)
train = prepare_dataset(train)
test = prepare_dataset(test)
logging.info('> Exporting train dataset')
train.to_csv(input_folder/train_file)
logging.info('> Exporting test dataset')
test.to_csv(input_folder/test_file)
if args.check_dataset:
from generate_simulation_data import GenerateSimulationData as sim
print('\nChecking conformity of %s %s dataset…' % (d, c))
sim.check_dataset_conformity(run_dir,
run_img_dir,
'Dataset - %s avg_gap-%s %s' %
(args.net_input, args.avg_gap, c),
c,
net_input=args.net_input,
communication=communication)
if args.train_net or args.plots_net:
from utils.utils import prepare_dataset
indices = prepare_dataset(run_dir, args.generate_split,
args.n_simulations)
file_losses = os.path.join(model_dir, 'losses.npy')
if args.train_net:
from network_training_distribute import network_train
network_train(indices,
file_losses,
runs_dir_omniscient,
model_dir,
args.model,
communication,
net_input=args.net_input,
save_net=args.save_net)
if args.plots_net:
from network_evaluation import network_evaluation
def reconst_loss(self, inputs):
''' ------------------------------------------------------------------------------
DATA PROCESSING
------------------------------------------------------------------------------ '''
inputs = utils.prepare_dataset(inputs)
return self.model.reconst_loss(inputs)
def interpolate(self, input1, input2):
input1 = utils.prepare_dataset(input1)
input2 = utils.prepare_dataset(input2)
return self.model.interpolate(input1, input2)
def encode(self, inputs):
''' ------------------------------------------------------------------------------
DATA PROCESSING
------------------------------------------------------------------------------ '''
inputs = utils.prepare_dataset(inputs)
return self.model.encode(inputs)