def validateProvideBatch_TFRecord(self, truncated_length, batch_size,
lengths, expected_num_inputs):
hparams = copy.deepcopy(constants.DEFAULT_HPARAMS)
examples = []
expected_inputs = []
for i, length in enumerate(lengths):
wav_samples = np.zeros(
(np.int((length / data.hparams_frames_per_second(hparams)) *
constants.DEFAULT_SAMPLE_RATE), 1), np.float32)
wav_data = audio_io.samples_to_wav_data(
wav_samples, constants.DEFAULT_SAMPLE_RATE)
num_frames = data.wav_to_num_frames(
wav_data,
frames_per_second=data.hparams_frames_per_second(hparams))
seq = self._SyntheticSequence(
num_frames / data.hparams_frames_per_second(hparams),
i + constants.MIN_MIDI_PITCH)
examples.append(self._FillExample(seq, wav_data, 'ex%d' % i))
expected_inputs += self._ExampleToInputs(examples[-1],
truncated_length)
self.assertEqual(expected_num_inputs, len(expected_inputs))
with tempfile.NamedTemporaryFile() as temp_rio:
with tf.python_io.TFRecordWriter(temp_rio.name) as writer:
for ex in examples:
writer.write(ex.SerializeToString())
self.validateProvideBatch(temp_rio.name, truncated_length,
batch_size, expected_inputs)
def _CreateExamplesAndExpectedInputs(self,
truncated_length,
lengths,
expected_num_inputs):
hparams = copy.deepcopy(configs.DEFAULT_HPARAMS)
examples = []
expected_inputs = []
for i, length in enumerate(lengths):
wav_samples = np.zeros(
(np.int((length / data.hparams_frames_per_second(hparams)) *
hparams.sample_rate), 1), np.float32)
wav_data = audio_io.samples_to_wav_data(wav_samples, hparams.sample_rate)
num_frames = data.wav_to_num_frames(
wav_data, frames_per_second=data.hparams_frames_per_second(hparams))
seq = self._SyntheticSequence(
num_frames / data.hparams_frames_per_second(hparams),
i + constants.MIN_MIDI_PITCH)
examples.append(self._FillExample(seq, wav_data, 'ex%d' % i))
expected_inputs += self._ExampleToInputs(
examples[-1],
truncated_length)
self.assertEqual(expected_num_inputs, len(expected_inputs))
return examples, expected_inputs
def _ExampleToInputs(self,
ex,
truncated_length=0):
hparams = copy.deepcopy(configs.DEFAULT_HPARAMS)
filename = ex.features.feature['id'].bytes_list.value[0]
sequence = music_pb2.NoteSequence.FromString(
ex.features.feature['sequence'].bytes_list.value[0])
wav_data = ex.features.feature['audio'].bytes_list.value[0]
spec = data.wav_to_spec(wav_data, hparams=hparams)
roll = sequences_lib.sequence_to_pianoroll(
sequence,
frames_per_second=data.hparams_frames_per_second(hparams),
min_pitch=constants.MIN_MIDI_PITCH,
max_pitch=constants.MAX_MIDI_PITCH,
min_frame_occupancy_for_label=0.0,
onset_mode='length_ms',
onset_length_ms=32.,
onset_delay_ms=0.)
length = data.wav_to_num_frames(
wav_data, frames_per_second=data.hparams_frames_per_second(hparams))
return self._DataToInputs(spec, roll.active, roll.weights, length, filename,
truncated_length)
def _ExampleToInputs(self,
ex,
truncated_length=0):
hparams = copy.deepcopy(configs.DEFAULT_HPARAMS)
filename = ex.features.feature['id'].bytes_list.value[0]
sequence = music_pb2.NoteSequence.FromString(
ex.features.feature['sequence'].bytes_list.value[0])
wav_data = ex.features.feature['audio'].bytes_list.value[0]
spec = data.wav_to_spec(wav_data, hparams=hparams)
roll = sequences_lib.sequence_to_pianoroll(
sequence,
frames_per_second=data.hparams_frames_per_second(hparams),
min_pitch=constants.MIN_MIDI_PITCH,
max_pitch=constants.MAX_MIDI_PITCH,
min_frame_occupancy_for_label=0.0,
onset_mode='length_ms',
onset_length_ms=32.,
onset_delay_ms=0.)
length = data.wav_to_num_frames(
wav_data, frames_per_second=data.hparams_frames_per_second(hparams))
return self._DataToInputs(spec, roll.active, roll.weights, length, filename,
truncated_length)
def _ExampleToInputs(self,
ex,
truncated_length=0,
crop_training_sequence_to_notes=False):
hparams = copy.deepcopy(constants.DEFAULT_HPARAMS)
hparams.crop_training_sequence_to_notes = crop_training_sequence_to_notes
filename = ex.features.feature['id'].bytes_list.value[0]
sequence, crop_beginning_seconds = data.preprocess_sequence(
ex.features.feature['sequence'].bytes_list.value[0], hparams)
wav_data = ex.features.feature['audio'].bytes_list.value[0]
if crop_training_sequence_to_notes:
wav_data = audio_io.crop_wav_data(wav_data, hparams.sample_rate,
crop_beginning_seconds,
sequence.total_time)
spec = data.wav_to_spec(wav_data, hparams=hparams)
roll = sequences_lib.sequence_to_pianoroll(
sequence,
frames_per_second=data.hparams_frames_per_second(hparams),
min_pitch=constants.MIN_MIDI_PITCH,
max_pitch=constants.MAX_MIDI_PITCH,
min_frame_occupancy_for_label=0.0,
onset_mode='length_ms',
onset_length_ms=32.,
onset_delay_ms=0.)
length = data.wav_to_num_frames(
wav_data, frames_per_second=data.hparams_frames_per_second(hparams))
return self._DataToInputs(spec, roll.active, roll.weights, length, filename,
truncated_length)
def _CreateExamplesAndExpectedInputs(self,
truncated_length,
lengths,
expected_num_inputs):
hparams = copy.deepcopy(configs.DEFAULT_HPARAMS)
examples = []
expected_inputs = []
for i, length in enumerate(lengths):
wav_samples = np.zeros(
(np.int((length / data.hparams_frames_per_second(hparams)) *
hparams.sample_rate), 1), np.float32)
wav_data = audio_io.samples_to_wav_data(wav_samples, hparams.sample_rate)
num_frames = data.wav_to_num_frames(
wav_data, frames_per_second=data.hparams_frames_per_second(hparams))
seq = self._SyntheticSequence(
num_frames / data.hparams_frames_per_second(hparams),
i + constants.MIN_MIDI_PITCH)
examples.append(self._FillExample(seq, wav_data, 'ex%d' % i))
expected_inputs += self._ExampleToInputs(
examples[-1],
truncated_length)
self.assertEqual(expected_num_inputs, len(expected_inputs))
return examples, expected_inputs
def _ExampleToInputs(self,
ex,
truncated_length=0,
crop_training_sequence_to_notes=False):
hparams = copy.deepcopy(constants.DEFAULT_HPARAMS)
hparams.crop_training_sequence_to_notes = crop_training_sequence_to_notes
filename = ex.features.feature['id'].bytes_list.value[0]
sequence, crop_beginning_seconds = data.preprocess_sequence(
ex.features.feature['sequence'].bytes_list.value[0], hparams)
wav_data = ex.features.feature['audio'].bytes_list.value[0]
if crop_training_sequence_to_notes:
wav_data = audio_io.crop_wav_data(wav_data, hparams.sample_rate,
crop_beginning_seconds,
sequence.total_time)
spec = data.wav_to_spec(wav_data, hparams=hparams)
roll = sequences_lib.sequence_to_pianoroll(
sequence,
frames_per_second=data.hparams_frames_per_second(hparams),
min_pitch=constants.MIN_MIDI_PITCH,
max_pitch=constants.MAX_MIDI_PITCH,
min_frame_occupancy_for_label=0.0,
onset_mode='length_ms',
onset_length_ms=32.,
onset_delay_ms=0.)
length = data.wav_to_num_frames(
wav_data,
frames_per_second=data.hparams_frames_per_second(hparams))
return self._DataToInputs(spec, roll.active, roll.weights, length,
filename, truncated_length)
def _ValidateProvideBatchMemory(self,
truncated_length,
batch_size,
lengths,
expected_num_inputs):
hparams = copy.deepcopy(constants.DEFAULT_HPARAMS)
examples = []
expected_inputs = []
for i, length in enumerate(lengths):
wav_samples = np.zeros(
(np.int((length / data.hparams_frames_per_second(hparams)) *
hparams.sample_rate), 1), np.float32)
wav_data = audio_io.samples_to_wav_data(wav_samples, hparams.sample_rate)
num_frames = data.wav_to_num_frames(
wav_data, frames_per_second=data.hparams_frames_per_second(hparams))
seq = self._SyntheticSequence(
num_frames / data.hparams_frames_per_second(hparams),
i + constants.MIN_MIDI_PITCH)
examples.append(self._FillExample(seq, wav_data, 'ex%d' % i))
expected_inputs += self._ExampleToInputs(
examples[-1],
truncated_length)
self.assertEqual(expected_num_inputs, len(expected_inputs))
self._ValidateProvideBatch(
[e.SerializeToString() for e in examples],
truncated_length,
batch_size,
expected_inputs)
def _ValidateProvideBatchMemory(self, truncated_length, batch_size,
lengths, expected_num_inputs):
hparams = copy.deepcopy(constants.DEFAULT_HPARAMS)
examples = []
expected_inputs = []
for i, length in enumerate(lengths):
wav_samples = np.zeros(
(np.int((length / data.hparams_frames_per_second(hparams)) *
hparams.sample_rate), 1), np.float32)
wav_data = audio_io.samples_to_wav_data(wav_samples,
hparams.sample_rate)
num_frames = data.wav_to_num_frames(
wav_data,
frames_per_second=data.hparams_frames_per_second(hparams))
seq = self._SyntheticSequence(
num_frames / data.hparams_frames_per_second(hparams),
i + constants.MIN_MIDI_PITCH)
examples.append(self._FillExample(seq, wav_data, 'ex%d' % i))
expected_inputs += self._ExampleToInputs(examples[-1],
truncated_length)
self.assertEqual(expected_num_inputs, len(expected_inputs))
self._ValidateProvideBatch([e.SerializeToString() for e in examples],
truncated_length, batch_size,
expected_inputs)
def _ValidateProvideBatchTFRecord(self,
truncated_length,
batch_size,
lengths,
expected_num_inputs,
crop_sequence_secs=0):
hparams = copy.deepcopy(constants.DEFAULT_HPARAMS)
examples = []
expected_inputs = []
for i, length in enumerate(lengths):
wav_samples = np.zeros(
(np.int((length / data.hparams_frames_per_second(hparams)) *
hparams.sample_rate), 1), np.float32)
wav_data = audio_io.samples_to_wav_data(wav_samples,
hparams.sample_rate)
num_frames = data.wav_to_num_frames(
wav_data,
frames_per_second=data.hparams_frames_per_second(hparams))
seq = self._SyntheticSequence(
num_frames / data.hparams_frames_per_second(hparams) -
crop_sequence_secs * 2, # crop from both ends.
i + constants.MIN_MIDI_PITCH,
start_time=crop_sequence_secs)
examples.append(self._FillExample(seq, wav_data, 'ex%d' % i))
expected_inputs += self._ExampleToInputs(
examples[-1],
truncated_length,
crop_training_sequence_to_notes=crop_sequence_secs > 0)
self.assertEqual(expected_num_inputs, len(expected_inputs))
with tempfile.NamedTemporaryFile() as temp_tfr:
with tf.python_io.TFRecordWriter(temp_tfr.name) as writer:
for ex in examples:
writer.write(ex.SerializeToString())
self._ValidateProvideBatch(
temp_tfr.name,
truncated_length,
batch_size,
expected_inputs,
crop_training_sequence_to_notes=crop_sequence_secs > 0)
def _ValidateProvideBatchTFRecord(self,
truncated_length,
batch_size,
lengths,
expected_num_inputs,
crop_sequence_secs=0):
hparams = copy.deepcopy(constants.DEFAULT_HPARAMS)
examples = []
expected_inputs = []
for i, length in enumerate(lengths):
wav_samples = np.zeros(
(np.int((length / data.hparams_frames_per_second(hparams)) *
hparams.sample_rate), 1), np.float32)
wav_data = audio_io.samples_to_wav_data(wav_samples, hparams.sample_rate)
num_frames = data.wav_to_num_frames(
wav_data, frames_per_second=data.hparams_frames_per_second(hparams))
seq = self._SyntheticSequence(
num_frames / data.hparams_frames_per_second(hparams) -
crop_sequence_secs * 2, # crop from both ends.
i + constants.MIN_MIDI_PITCH,
start_time=crop_sequence_secs)
examples.append(self._FillExample(seq, wav_data, 'ex%d' % i))
expected_inputs += self._ExampleToInputs(
examples[-1],
truncated_length,
crop_training_sequence_to_notes=crop_sequence_secs > 0)
self.assertEqual(expected_num_inputs, len(expected_inputs))
with tempfile.NamedTemporaryFile() as temp_tfr:
with tf.python_io.TFRecordWriter(temp_tfr.name) as writer:
for ex in examples:
writer.write(ex.SerializeToString())
self._ValidateProvideBatch(
temp_tfr.name,
truncated_length,
batch_size,
expected_inputs,
crop_training_sequence_to_notes=crop_sequence_secs > 0)
def validateProvideBatch_TFRecord(self,
truncated_length,
batch_size,
lengths,
expected_num_inputs):
hparams = copy.deepcopy(constants.DEFAULT_HPARAMS)
examples = []
expected_inputs = []
for i, length in enumerate(lengths):
wav_samples = np.zeros(
(np.int((length / data.hparams_frames_per_second(hparams)) *
constants.DEFAULT_SAMPLE_RATE), 1), np.float32)
wav_data = audio_io.samples_to_wav_data(wav_samples,
constants.DEFAULT_SAMPLE_RATE)
num_frames = data.wav_to_num_frames(
wav_data, frames_per_second=data.hparams_frames_per_second(hparams))
seq = self._SyntheticSequence(
num_frames / data.hparams_frames_per_second(hparams),
i + constants.MIN_MIDI_PITCH)
examples.append(self._FillExample(seq, wav_data, 'ex%d' % i))
expected_inputs += self._ExampleToInputs(
examples[-1],
truncated_length)
self.assertEqual(expected_num_inputs, len(expected_inputs))
with tempfile.NamedTemporaryFile() as temp_rio:
with tf.python_io.TFRecordWriter(temp_rio.name) as writer:
for ex in examples:
writer.write(ex.SerializeToString())
self.validateProvideBatch(
temp_rio.name,
truncated_length,
batch_size,
expected_inputs)
