def initialize(self):
# create save directory
if os.path.exists(self.save_dir):
eprint('Confirm removing {}\n[Y/n]'.format(self.save_dir))
if input() != 'Y':
import sys
sys.exit()
import shutil
shutil.rmtree(self.save_dir)
eprint('Removed: ' + self.save_dir)
os.makedirs(self.save_dir)
# set deterministic random seed
if self.random_seed is not None:
reset_random(self.random_seed)
def initialize(cls, config):
# create save directory
if os.path.exists(config.save_dir):
eprint('Confirm removing {}\n[Y/n]'.format(config.save_dir))
_input = input()
if _input == 'Y':
import shutil
shutil.rmtree(config.save_dir)
eprint('Removed: ' + config.save_dir)
elif _input != 'n':
import sys
sys.exit()
if not os.path.exists(config.save_dir):
os.makedirs(config.save_dir)
# set deterministic random seed
if config.random_seed is not None:
reset_random(config.random_seed)
def initialize(self):
# arXiv 1509.09308
# a new class of fast algorithms for convolutional neural networks using Winograd's minimal filtering algorithms
os.environ['TF_ENABLE_WINOGRAD_NONFUSED'] = '1'
# create training directory
if not self.restore:
if os.path.exists(self.train_dir):
eprint('Confirm removing {}\n[Y/n]'.format(self.train_dir))
if input() != 'Y':
import sys
sys.exit()
import shutil
shutil.rmtree(self.train_dir, ignore_errors=True)
eprint('Removed: ' + self.train_dir)
os.makedirs(self.train_dir)
# set deterministic random seed
if self.random_seed is not None:
reset_random(self.random_seed)
def initialize(self):
if not os.path.exists(self.output_dir):
os.makedirs(self.output_dir)
# set deterministic random seed
if self.random_seed is not None:
reset_random(self.random_seed)