def main(_):
raw_data = reader.load_holmes_data(12001)
train_data, _, word_to_id = raw_data
if FLAGS.model == "test":
test_questions = reader.get_questions(word_to_id, FLAGS.q_path)
config = get_config()
eval_config = get_config()
eval_config.batch_size = 1
eval_config.num_steps = 30
with tf.Graph().as_default():
initializer = tf.random_uniform_initializer(-config.init_scale,
config.init_scale)
with tf.name_scope("Train"):
train_input = PTBInput(config=config,
data=train_data,
name="TrainInput")
with tf.variable_scope("Model",
reuse=None,
initializer=initializer):
m = PTBModel(is_training=True,
config=config,
input_=train_input)
tf.summary.scalar("Training Loss", m.cost)
tf.summary.scalar("Learning Rate", m.lr)
if FLAGS.model == "test":
with tf.name_scope("Test"):
test_input = PTBInput(config=eval_config,
data=train_data,
name="TestInput")
with tf.variable_scope("Model",
reuse=True,
initializer=initializer):
mtest = PTBModel(is_training=False,
config=eval_config,
input_=test_questions)
saver = tf.train.Saver()
session_config = tf.ConfigProto()
session_config.gpu_options.per_process_gpu_memory_fraction = 0.05
with tf.Session(config=session_config) as session:
session.run(initializer)
for i in range(config.max_max_epoch):
lr_decay = config.lr_decay**max(i + 1 - config.max_epoch, 0.0)
m.assign_lr(session, config.learning_rate * lr_decay)
print("Epoch: %d Learning rate: %.3f" %
(i + 1, session.run(m.lr)))
train_perplexity = run_epoch(session,
m,
eval_op=m.train_op,
verbose=True)
save_path = saver.save(session, "model/model.ckpt")
print("Model saved in file: %s" % save_path)
saver.restore(session, "model/model.ckpt")
print("Model restored.")
test_perplexity = predict(session, mtest, test_questions)
def cli(question_file, answers_dir, answer_file, test):
"""Given a set of questions and the answers directory with top 100
documents for each question, generate the answer file
"""
success('---NLP Project Three: Question Answer---')
question_filepath = os.path.realpath(question_file)
answers_dirpath = os.path.realpath(answers_dir)
answer_filepath = os.path.realpath(answer_file)
log('Answering: {}\n Using: {}\n Into: {}'.format(question_filepath,
answers_dirpath,
answer_filepath))
if test:
warn('Testing, not normal execution...')
_test_endpoint(question_filepath, answers_dirpath, answer_filepath)
return
try:
questions = get_questions(question_filepath)
if len(questions) is not 232:
warn('devset has 232 questions (Got {})'.format(len(questions)))
answers = {}
for question in questions:
answers[question.qid] = get_answers(question, answers_dirpath)
if len(answers) is not len(questions):
warn('Got {} answers for {} questions'.format(
len(answers), len(questions)))
write_answers(answers, answer_filepath)
success('Wrote answers to {}'.format(answer_filepath))
except NotImplementedError as e:
error('TODO: {}'.format(e))
def _test_endpoint(question_filepath, answers_dirpath, answer_filepath):
"""Testing endpoint
"""
questions = get_questions(question_filepath)
print questions
question = [q for q in questions if q.qid == 214]
print question
question = question[0]
answers = get_answers(question, answers_dirpath)
print answers