class NotePerformanceEventSequenceEncoderDecoderTest(tf.test.TestCase):
def setUp(self):
self.enc = NotePerformanceEventSequenceEncoderDecoder(
num_velocity_bins=16, max_shift_steps=99, max_duration_steps=500)
self.assertEqual(10, self.enc.shift_steps_segments)
self.assertEqual(20, self.enc.duration_steps_segments)
def testEncodeDecode(self):
pe = PerformanceEvent
performance = [
(pe(pe.TIME_SHIFT, 0), pe(pe.NOTE_ON,
60), pe(pe.VELOCITY,
13), pe(pe.DURATION, 401)),
(pe(pe.TIME_SHIFT, 55), pe(pe.NOTE_ON,
64), pe(pe.VELOCITY,
13), pe(pe.DURATION, 310)),
(pe(pe.TIME_SHIFT, 99), pe(pe.NOTE_ON,
67), pe(pe.VELOCITY,
16), pe(pe.DURATION, 100)),
(pe(pe.TIME_SHIFT, 0), pe(pe.NOTE_ON,
67), pe(pe.VELOCITY,
16), pe(pe.DURATION, 1)),
(pe(pe.TIME_SHIFT, 0), pe(pe.NOTE_ON,
67), pe(pe.VELOCITY,
16), pe(pe.DURATION, 500)),
]
labels = [
self.enc.events_to_label(performance, i)
for i in range(len(performance))
]
expected_labels = [
(0, 0, 60, 12, 16, 0),
(5, 5, 64, 12, 12, 9),
(9, 9, 67, 15, 3, 24),
(0, 0, 67, 15, 0, 0),
(0, 0, 67, 15, 19, 24),
]
self.assertEqual(expected_labels, labels)
inputs = [
self.enc.events_to_input(performance, i)
for i in range(len(performance))
]
for input_ in inputs:
self.assertEqual(6, input_.nonzero()[0].shape[0])
decoded_performance = [
self.enc.class_index_to_event(label, None) for label in labels
]
self.assertEqual(performance, decoded_performance)
class NotePerformanceEventSequenceEncoderDecoderTest(tf.test.TestCase):
def setUp(self):
self.enc = NotePerformanceEventSequenceEncoderDecoder(
num_velocity_bins=16, max_shift_steps=99, max_duration_steps=500)
self.assertEqual(10, self.enc.shift_steps_segments)
self.assertEqual(20, self.enc.duration_steps_segments)
def testEncodeDecode(self):
pe = PerformanceEvent
performance = [
(pe(pe.TIME_SHIFT, 0), pe(pe.NOTE_ON, 60),
pe(pe.VELOCITY, 13), pe(pe.DURATION, 401)),
(pe(pe.TIME_SHIFT, 55), pe(pe.NOTE_ON, 64),
pe(pe.VELOCITY, 13), pe(pe.DURATION, 310)),
(pe(pe.TIME_SHIFT, 99), pe(pe.NOTE_ON, 67),
pe(pe.VELOCITY, 16), pe(pe.DURATION, 100)),
(pe(pe.TIME_SHIFT, 0), pe(pe.NOTE_ON, 67),
pe(pe.VELOCITY, 16), pe(pe.DURATION, 1)),
(pe(pe.TIME_SHIFT, 0), pe(pe.NOTE_ON, 67),
pe(pe.VELOCITY, 16), pe(pe.DURATION, 500)),
]
labels = [self.enc.events_to_label(performance, i)
for i in range(len(performance))]
expected_labels = [
(0, 0, 60, 12, 16, 0),
(5, 5, 64, 12, 12, 9),
(9, 9, 67, 15, 3, 24),
(0, 0, 67, 15, 0, 0),
(0, 0, 67, 15, 19, 24),
]
self.assertEqual(expected_labels, labels)
inputs = [self.enc.events_to_input(performance, i)
for i in range(len(performance))]
for input_ in inputs:
self.assertEqual(6, input_.nonzero()[0].shape[0])
decoded_performance = [self.enc.class_index_to_event(label, None)
for label in labels]
self.assertEqual(performance, decoded_performance)
def setUp(self):
self.enc = NotePerformanceEventSequenceEncoderDecoder(
num_velocity_bins=16, max_shift_steps=99, max_duration_steps=500)
self.assertEqual(10, self.enc.shift_steps_segments)
self.assertEqual(20, self.enc.duration_steps_segments)
def setUp(self):
self.enc = NotePerformanceEventSequenceEncoderDecoder(
num_velocity_bins=16, max_shift_steps=99, max_duration_steps=500)
self.assertEqual(10, self.enc.shift_steps_segments)
self.assertEqual(20, self.enc.duration_steps_segments)