import argparse
from parserr import Parser
from datamanager import DataManager
from actor import ActorNetwork
from LSTM_critic import LSTM_CriticNetwork
tf.logging.set_verbosity(tf.logging.ERROR)
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
#get parse
argv = sys.argv[1:]
parser = Parser().getParser()
args, _ = parser.parse_known_args(argv)
random.seed(args.seed)
#get data
dataManager = DataManager(args.dataset)
train_data, dev_data, test_data = dataManager.getdata(args.grained, args.maxlenth)
word_vector = dataManager.get_wordvector(args.word_vector)
if args.fasttest == 1:
train_data = train_data[:100]
dev_data = dev_data[:20]
test_data = test_data[:20]
print("train_data ", len(train_data))
print("dev_data", len(dev_data))
print("test_data", len(test_data))
def sampling_RL(sess, actor, inputs, vec, lenth, Random=True):
current_lower_state = np.zeros((1, 2*args.dim), dtype=np.float32)
actions = []
states = []
#sampling actions
#get parse
argv = sys.argv[1:]
parser = Parser().getParser()
args, _ = parser.parse_known_args(argv)
random.seed(args.seed)
#get data
ME_DIR = os.path.dirname(os.path.realpath(__file__))
work_dir = ME_DIR
embedding_file = work_dir + '/embedding/glove.twitter.27B.200d.txt'
emoji_embedding_file = work_dir + '/embedding/emoji2vec.txt'
embedding_file_ = work_dir+'/embedding/dict_file.csv'
embedding_dim = 200
datamanager = DataManager('a')
train_data, test_data, dev_data = datamanager.getdata(2, args.maxlenth)
word_vector = datamanager.get_wordvector(embedding_file,emoji_embedding_file)
def sampling_RL(sess, actor, inputs, lenth, Random=True):
current_lower_state = np.zeros((1, state_size), dtype=np.float32)
current_upper_state = np.zeros((1, state_size), dtype=np.float32)
actions = []
states = []
#sampling actions
for pos in range(lenth):
out_d, current_lower_state = critic.lower_LSTM_target(current_lower_state, [[inputs[pos]]])
predicted = actor.predict_target(current_upper_state, current_lower_state)
#print predicted
states.append([current_upper_state, current_lower_state])
if Random: