def construct_dir(self): self.model_dir = utils.construct_dir(prefix=self.args.model_dir, args=self.args, create_dataset_name=False) self.summary_dir = utils.construct_dir(prefix=self.args.summary_dir, args=self.args, create_dataset_name=False) self.out_path = utils.construct_dir(prefix=self.args.result_dir, args=self.args, create_dataset_name=False) + '.txt' if not os.path.exists(self.args.result_dir): os.makedirs(self.args.result_dir) if not os.path.exists(self.model_dir): os.makedirs(self.model_dir) if not os.path.exists(self.summary_dir): os.makedirs(self.summary_dir)
def construct_out_dir(self):
self.result_dir = utils.construct_dir(prefix=self.args.result_dir,
args=self.args)
self.model_dir = os.path.join(self.result_dir, 'models')
self.out_path = os.path.join(self.result_dir, 'result.txt')
self.summary_dir = utils.construct_dir(prefix=self.args.summary_dir,
args=self.args)
self.image_dir = utils.construct_dir(prefix='images', args=self.args)
if not os.path.exists(self.summary_dir):
os.makedirs(self.summary_dir)
self.writer = SummaryWriter(log_dir=self.summary_dir)
if not os.path.exists(self.model_dir):
os.makedirs(self.model_dir)
def construct_data(self):
self.data_dir = os.path.join(self.args.data_dir, self.args.dataset)
self.dataset_name = utils.construct_dir(prefix=self.args.dataset, args=self.args, create_dataset_name=True)
dataset_file_name = os.path.join(self.data_dir, self.dataset_name)
if not os.path.exists(dataset_file_name):
if self.args.dataset == 'imdb':
self.text_data = IMDBData(args=self.args)
elif self.args.dataset == 'agnews':
self.text_data = AGNewsData(args=self.args)
else:
print('Cannot recognize {}'.format(self.args.dataset))
raise NotImplementedError
with open(dataset_file_name, 'wb') as datasetFile:
p.dump(self.text_data, datasetFile)
print('dataset created and saved to {}, exiting ...'.format(dataset_file_name))
exit(0)
else:
with open(dataset_file_name, 'rb') as datasetFile:
self.text_data = p.load(datasetFile)
print('dataset loaded from {}'.format(dataset_file_name))
# construct dataset
self.text_data.construct_dataset(max_steps=self.args.max_steps)
self.train_loader = DataLoader(dataset=self.text_data.training_set_all, num_workers=self.args.num_worker, batch_size=self.args.batch_size, shuffle=True)
self.val_loader = DataLoader(dataset=self.text_data.val_set, num_workers=self.args.num_worker, batch_size=self.args.batch_size, shuffle=False)
self.test_loader = DataLoader(dataset=self.text_data.test_set, num_workers=self.args.num_worker, batch_size=self.args.batch_size, shuffle=False)
def construct_out_dir(self):
self.model_dir = utils.construct_dir(prefix=self.args.model_dir,
args=self.args,
create_dataset_name=False)
self.out_dir = utils.construct_dir(prefix=self.args.test_dir,
args=self.args,
create_dataset_name=False)
self.result_file = self.model_dir.split('/')[-1]
self.out_path = os.path.join(self.args.result_dir, self.result_file)
if not os.path.exists(self.args.result_dir):
os.makedirs(self.args.result_dir)
if not os.path.exists(self.model_dir):
os.makedirs(self.model_dir)
if not os.path.exists(self.out_dir):
os.makedirs(self.out_dir)
def construct_out_dir(self):
self.result_dir = utils.construct_dir(prefix=self.args.result_dir,
args=self.args)
self.out_image_dir = os.path.join(self.result_dir, 'images')
self.model_dir = os.path.join(self.result_dir, 'models')
self.out_path = os.path.join(self.result_dir, 'result.txt')
if not os.path.exists(self.out_image_dir):
os.makedirs(self.out_image_dir)
if not os.path.exists(self.model_dir):
os.makedirs(self.model_dir)
def construct_data(self):
self.data_dir = os.path.join(self.args.data_dir, self.args.dataset)
self.dataset_name = utils.construct_dir(prefix=self.args.dataset,
args=self.args,
create_dataset_name=True)
dataset_file_name = os.path.join(self.data_dir, self.dataset_name)
if not os.path.exists(dataset_file_name):
if self.args.dataset == 'rotten':
self.text_data = RottenData(args=self.args)
elif self.args.dataset == 'congress':
self.text_data = CongressData(args=self.args)
else:
print('Cannot recognize {}'.format(self.args.dataset))
raise NotImplementedError
with open(dataset_file_name, 'wb') as datasetFile:
p.dump(self.text_data, datasetFile)
print('dataset created and saved to {}, exiting ...'.format(
dataset_file_name))
exit(0)
else:
with open(dataset_file_name, 'rb') as datasetFile:
self.text_data = p.load(datasetFile)
print('dataset loaded from {}'.format(dataset_file_name))
self.text_data.construct_dataset(elmo=self.args.elmo)
self.train_loader = DataLoader(dataset=self.text_data.training_dataset,
num_workers=1,
batch_size=self.args.batch_size,
shuffle=False)
self.val_loader = DataLoader(dataset=self.text_data.val_dataset,
num_workers=1,
batch_size=self.args.test_batch_size,
shuffle=False)
self.test_loader = DataLoader(dataset=self.text_data.test_dataset,
num_workers=1,
batch_size=self.args.test_batch_size,
shuffle=False)
def compute_dist(self): self.vocab = self.text_data.get_vocab() dists = compute_dist(vocab=self.vocab, word2id=self.text_data.word2id, id2word=self.text_data.id2word, embedding_file='counter-fitted-vectors.txt') dist_name = utils.construct_dir(prefix=self.args.dataset, args=self.args, create_dist_name=True) with open(dist_name, 'wb') as file: p.dump(dists, file)
