def run(melody_encoder_decoder):
"""Creates training and eval data with the given MelodyEncoderDecoder.
Args:
melody_encoder_decoder: A melodies_lib.MelodyEncoderDecoder.
"""
tf.logging.set_verbosity(tf.logging.INFO)
if not FLAGS.input:
tf.logging.fatal('--input required')
return
if not FLAGS.train_output:
tf.logging.fatal('--train_output required')
return
FLAGS.input = os.path.expanduser(FLAGS.input)
FLAGS.train_output = os.path.expanduser(FLAGS.train_output)
if FLAGS.eval_output:
FLAGS.eval_output = os.path.expanduser(FLAGS.eval_output)
if not os.path.exists(os.path.dirname(FLAGS.train_output)):
os.makedirs(os.path.dirname(FLAGS.train_output))
if FLAGS.eval_output:
if not os.path.exists(os.path.dirname(FLAGS.eval_output)):
os.makedirs(os.path.dirname(FLAGS.eval_output))
sequence_to_melodies.run_conversion(melody_encoder_decoder,
FLAGS.input,
FLAGS.train_output,
FLAGS.eval_output,
FLAGS.eval_ratio)
def testRunConversionNoEval(self):
sequence_to_melodies.run_conversion(
encoder=sequence_to_melodies.basic_one_hot_encoder,
sequences_file=self.sequences_file,
train_output=self.train_output)
self.assertTrue(os.path.isfile(self.train_output))
reader = tf.python_io.tf_record_iterator(self.train_output)
self.assertEqual(62, len(list(reader)))
self.assertFalse(os.path.isfile(self.eval_output))
def main(unused_argv):
root = logging.getLogger()
root.setLevel(logging.INFO)
ch = logging.StreamHandler(sys.stdout)
ch.setLevel(logging.INFO)
root.addHandler(ch)
sequence_to_melodies.run_conversion(
encoder=sequence_to_melodies.basic_one_hot_encoder,
sequences_file=FLAGS.input,
train_output=FLAGS.train_output, eval_output=FLAGS.eval_output,
eval_ratio=FLAGS.eval_ratio)
def testRunConversion(self):
sequence_to_melodies.run_conversion(
encoder=sequence_to_melodies.basic_one_hot_encoder,
sequences_file=self.sequences_file,
train_output=self.train_output, eval_output=self.eval_output,
eval_ratio=0.25)
self.assertTrue(os.path.isfile(self.train_output))
reader = tf.python_io.tf_record_iterator(self.train_output)
num_train_samples = len(list(reader))
self.assertTrue(os.path.isfile(self.eval_output))
reader = tf.python_io.tf_record_iterator(self.eval_output)
num_eval_samples = len(list(reader))
self.assertTrue(num_train_samples > 0)
self.assertTrue(num_eval_samples > 0)
self.assertTrue(num_train_samples > num_eval_samples)