def start_session(self):
""" Launch the tensorflow session and start the GuessWhat loop """
with tf.Session(config=self.tf_config) as sess:
guesser_network = GuesserNetwork(self.guesser_config['model'],
num_words=self.tokenizer.no_words)
guesser_var = [
v for v in tf.global_variables() if 'guesser' in v.name
]
guesser_saver = tf.train.Saver(var_list=guesser_var)
guesser_saver.restore(sess, GUESS_NTW_PATH)
guesser_wrapper = GuesserROSWrapper(guesser_network)
qgen_network = QGenNetworkLSTM(self.qgen_config['model'],
num_words=self.tokenizer.no_words,
policy_gradient=False)
qgen_var = [v for v in tf.global_variables() if 'qgen' in v.name]
qgen_saver = tf.train.Saver(var_list=qgen_var)
qgen_saver.restore(sess, QGEN_NTW_PATH)
qgen_network.build_sampling_graph(
self.qgen_config['model'],
tokenizer=self.tokenizer,
max_length=self.eval_config['loop']['max_depth'])
qgen_wrapper = QGenWrapper(
qgen_network,
self.tokenizer,
max_length=self.eval_config['loop']['max_depth'],
k_best=self.eval_config['loop']['beam_k_best'])
oracle_wrapper = OracleROSWrapper(self.tokenizer)
self.loop(sess, guesser_wrapper, qgen_wrapper, oracle_wrapper)
logger.info('Loading data..')
trainset = Dataset(args.data_dir, "train", image_loader, crop_loader)
validset = Dataset(args.data_dir, "valid", image_loader, crop_loader)
testset = Dataset(args.data_dir, "test", image_loader, crop_loader)
# Load dictionary
logger.info('Loading dictionary..')
tokenizer = GWTokenizer(os.path.join(args.data_dir, args.dict_file))
###############################
# LOAD NETWORKS
#############################
logger.info('Building networks..')
qgen_network = QGenNetworkLSTM(qgen_config["model"], num_words=tokenizer.no_words, policy_gradient=True)
qgen_var = [v for v in tf.global_variables() if "qgen" in v.name and 'rl_baseline' not in v.name]
qgen_saver = tf.train.Saver(var_list=qgen_var)
oracle_network = OracleNetwork(oracle_config, num_words=tokenizer.no_words)
oracle_var = [v for v in tf.global_variables() if "oracle" in v.name]
oracle_saver = tf.train.Saver(var_list=oracle_var)
guesser_network = GuesserNetwork(guesser_config["model"], num_words=tokenizer.no_words)
guesser_var = [v for v in tf.global_variables() if "guesser" in v.name]
guesser_saver = tf.train.Saver(var_list=guesser_var)
loop_saver = tf.train.Saver(allow_empty=False)
###############################
# REINFORCE OPTIMIZER
crop_loader = None # get_img_loader(config, 'crop', args.image_dir)
# Load data
logger.info('Loading data..')
trainset = Dataset(args.data_dir, "train", image_loader, crop_loader)
validset = Dataset(args.data_dir, "valid", image_loader, crop_loader)
testset = Dataset(args.data_dir, "test", image_loader, crop_loader)
# Load dictionary
logger.info('Loading dictionary..')
tokenizer = GWTokenizer(os.path.join(args.data_dir, args.dict_file))
# Build Network
logger.info('Building network..')
network = QGenNetworkLSTM(config["model"],
num_words=tokenizer.no_words,
policy_gradient=False)
# Build Optimizer
logger.info('Building optimizer..')
optimizer, outputs = create_optimizer(network, config)
###############################
# START TRAINING
#############################
# Load config
batch_size = config['optimizer']['batch_size']
no_epoch = config["optimizer"]["no_epoch"]
# create a saver to store/load checkpoint
os.path.join(args.networks_dir, 'guesser',
args.guesser_identifier, 'params.ckpt'))
guesser_wrapper = GuesserWrapper(guesser_network)
else:
guesser_wrapper = GuesserUserWrapper(tokenizer,
img_raw_dir=args.img_raw_dir)
logger.info("No Guesser was registered >>> use user input")
if args.qgen_identifier is not None:
qgen_config = get_config_from_xp(
os.path.join(args.networks_dir, "qgen"), args.qgen_identifier)
qgen_network = QGenNetworkLSTM(qgen_config["model"],
num_words=tokenizer.no_words,
policy_gradient=False)
qgen_var = [v for v in tf.global_variables()
if "qgen" in v.name] # and 'rl_baseline' not in v.name
qgen_saver = tf.train.Saver(var_list=qgen_var)
qgen_saver.restore(
sess,
os.path.join(args.networks_dir, 'qgen', args.qgen_identifier,
'params.ckpt'))
qgen_network.build_sampling_graph(
qgen_config["model"],
tokenizer=tokenizer,
max_length=eval_config['loop']['max_depth'])
qgen_wrapper = QGenWrapper(