def build_language_training_rnn(sess, hyper_params, prog_params, train_set, test_set):
model = LanguageModel(hyper_params["num_layers"], hyper_params["hidden_size"], hyper_params["batch_size"],
hyper_params["max_input_seq_length"], hyper_params["max_target_seq_length"],
hyper_params["char_map_length"])
# Create a Dataset from the train_set and the test_set
train_dataset = model.build_dataset(train_set, hyper_params["batch_size"], hyper_params["max_input_seq_length"],
hyper_params["char_map"])
v_iterator = None
if test_set is []:
t_iterator = model.add_dataset_input(train_dataset)
sess.run(t_iterator.initializer)
else:
test_dataset = model.build_dataset(test_set, hyper_params["batch_size"], hyper_params["max_input_seq_length"],
hyper_params["char_map"])
# Build the input stream from the different datasets
t_iterator, v_iterator = model.add_datasets_input(train_dataset, test_dataset)
sess.run(t_iterator.initializer)
sess.run(v_iterator.initializer)
# Create the model
model.create_training_rnn(hyper_params["dropout_input_keep_prob"], hyper_params["dropout_output_keep_prob"],
hyper_params["grad_clip"], hyper_params["learning_rate"],
hyper_params["lr_decay_factor"], use_iterator=True)
model.add_tensorboard(sess, hyper_params["tensorboard_dir"], prog_params["tb_name"], prog_params["timeline"])
model.initialize(sess)
model.restore(sess, hyper_params["checkpoint_dir"] + "/language/")
# Override the learning rate if given on the command line
if prog_params["learn_rate"] is not None:
model.set_learning_rate(sess, prog_params["learn_rate"])
return model, t_iterator, v_iterator
def generate_text(hyper_params):
with tf.Session() as sess:
# Create model
model = LanguageModel(hyper_params["num_layers"],
hyper_params["hidden_size"], 1, 1,
hyper_params["max_target_seq_length"],
hyper_params["char_map_length"])
model.create_forward_rnn()
model.initialize(sess)
model.restore(sess, hyper_params["checkpoint_dir"] + "/language/")
# Start with a letter
text = "O"
for _ in range(10):
print(text, end="")
# Convert to an one-hot encoded vector
input_vec = dataprocessor.DataProcessor.get_str_to_one_hot_encoded(
hyper_params["char_map"], text, add_eos=False)
feat_vec = np.array(input_vec)
(a, b) = feat_vec.shape
feat_vec = feat_vec.reshape((a, 1, b))
prediction = model.process_input(sess, feat_vec, [1])
text = dataprocessor.DataProcessor.get_labels_str(
hyper_params["char_map"], prediction[0])
print(text)
return
def build_language_training_rnn(sess, hyper_params, prog_params, train_set, test_set):
model = LanguageModel(hyper_params["num_layers"], hyper_params["hidden_size"], hyper_params["batch_size"],
hyper_params["max_input_seq_length"], hyper_params["max_target_seq_length"],
hyper_params["char_map_length"])
# Create a Dataset from the train_set and the test_set
train_dataset = model.build_dataset(train_set, hyper_params["batch_size"], hyper_params["max_input_seq_length"],
hyper_params["char_map"])
v_iterator = None
if test_set is []:
t_iterator = model.add_dataset_input(train_dataset)
sess.run(t_iterator.initializer)
else:
test_dataset = model.build_dataset(test_set, hyper_params["batch_size"], hyper_params["max_input_seq_length"],
hyper_params["char_map"])
# Build the input stream from the different datasets
t_iterator, v_iterator = model.add_datasets_input(train_dataset, test_dataset)
sess.run(t_iterator.initializer)
sess.run(v_iterator.initializer)
# Create the model
model.create_training_rnn(hyper_params["dropout_input_keep_prob"], hyper_params["dropout_output_keep_prob"],
hyper_params["grad_clip"], hyper_params["learning_rate"],
hyper_params["lr_decay_factor"], use_iterator=True)
model.add_tensorboard(sess, hyper_params["tensorboard_dir"], prog_params["tb_name"], prog_params["timeline"])
model.initialize(sess)
model.restore(sess, hyper_params["checkpoint_dir"] + "/language/")
# Override the learning rate if given on the command line
if prog_params["learn_rate"] is not None:
model.set_learning_rate(sess, prog_params["learn_rate"])
return model, t_iterator, v_iterator
def generate_text(hyper_params):
with tf.Session() as sess:
# Create model
model = LanguageModel(hyper_params["num_layers"], hyper_params["hidden_size"], 1, 1,
hyper_params["max_target_seq_length"], hyper_params["char_map_length"])
model.create_forward_rnn()
model.initialize(sess)
model.restore(sess, hyper_params["checkpoint_dir"] + "/language/")
# Start with a letter
text = "O"
for _ in range(10):
print(text, end="")
# Convert to an one-hot encoded vector
input_vec = dataprocessor.DataProcessor.get_str_to_one_hot_encoded(hyper_params["char_map"], text,
add_eos=False)
feat_vec = np.array(input_vec)
(a, b) = feat_vec.shape
feat_vec = feat_vec.reshape((a, 1, b))
prediction = model.process_input(sess, feat_vec, [1])
text = dataprocessor.DataProcessor.get_labels_str(hyper_params["char_map"], prediction[0])
print(text)
return