def __init__(self, checkpoint=None, bundle=None):
details = generator_pb2.GeneratorDetails(id='test_generator',
description='Test Generator')
super(SeuenceGenerator, self).__init__(Model(),
details,
checkpoint=checkpoint,
bundle=bundle)
def testUseMatchingGeneratorId(self):
bundle = generator_pb2.GeneratorBundle(
generator_details=generator_pb2.GeneratorDetails(
id='test_generator'),
checkpoint_file=[b'foo.ckpt'],
metagraph_file=b'foo.ckpt.meta')
SeuenceGenerator(bundle=bundle)
bundle.generator_details.id = 'blarg'
with self.assertRaises(sequence_generator.SequenceGeneratorError):
SeuenceGenerator(bundle=bundle)
def testSpecifyEitherCheckPointOrBundle(self):
bundle = generator_pb2.GeneratorBundle(
generator_details=generator_pb2.GeneratorDetails(
id='test_generator'),
checkpoint_file=[b'foo.ckpt'],
metagraph_file=b'foo.ckpt.meta')
with self.assertRaises(sequence_generator.SequenceGeneratorError):
SeuenceGenerator(checkpoint='foo.ckpt', bundle=bundle)
with self.assertRaises(sequence_generator.SequenceGeneratorError):
SeuenceGenerator(checkpoint=None, bundle=None)
SeuenceGenerator(checkpoint='foo.ckpt')
SeuenceGenerator(bundle=bundle)
def testGetBundleDetails(self):
# Test with non-bundle generator.
seq_gen = SeuenceGenerator(checkpoint='foo.ckpt')
self.assertEqual(None, seq_gen.bundle_details)
# Test with bundle-based generator.
bundle_details = generator_pb2.GeneratorBundle.BundleDetails(
description='bundle of joy')
bundle = generator_pb2.GeneratorBundle(
generator_details=generator_pb2.GeneratorDetails(
id='test_generator'),
bundle_details=bundle_details,
checkpoint_file=[b'foo.ckpt'],
metagraph_file=b'foo.ckpt.meta')
seq_gen = SeuenceGenerator(bundle=bundle)
self.assertEqual(bundle_details, seq_gen.bundle_details)
if (transpose_to_key is not None
and (transpose_to_key < 0
or transpose_to_key > mm.NOTES_PER_OCTAVE - 1)):
raise ValueError('transpose_to_key must be >= 0 and <= 11. '
'transpose_to_key is %d.' % transpose_to_key)
self.min_note = min_note
self.max_note = max_note
self.transpose_to_key = transpose_to_key
# Default configurations.
default_configs = {
'basic_rnn':
MelodyRnnConfig(
generator_pb2.GeneratorDetails(
id='basic_rnn', description='Melody RNN with one-hot encoding.'),
magenta.music.OneHotEventSequenceEncoderDecoder(
magenta.music.MelodyOneHotEncoding(min_note=DEFAULT_MIN_NOTE,
max_note=DEFAULT_MAX_NOTE)),
contrib_training.HParams(batch_size=128,
rnn_layer_sizes=[128, 128],
dropout_keep_prob=0.5,
clip_norm=5,
learning_rate=0.001)),
'mono_rnn':
MelodyRnnConfig(generator_pb2.GeneratorDetails(
id='mono_rnn', description='Monophonic RNN with one-hot encoding.'),
magenta.music.OneHotEventSequenceEncoderDecoder(
magenta.music.MelodyOneHotEncoding(min_note=0,
max_note=128)),
contrib_training.HParams(batch_size=128,
steps_per_iteration: The number of steps to take per beam search
iteration.
Returns:
The generated PianorollSequence object (which begins with the provided
primer track).
"""
return self._generate_events(
num_steps=num_steps, primer_events=primer_sequence, temperature=None,
beam_size=beam_size, branch_factor=branch_factor,
steps_per_iteration=steps_per_iteration)
default_configs = {
'rnn-nade': events_rnn_model.EventSequenceRnnConfig(
generator_pb2.GeneratorDetails(
id='rnn-nade',
description='RNN-NADE'),
mm.PianorollEncoderDecoder(),
contrib_training.HParams(
batch_size=64,
rnn_layer_sizes=[128, 128, 128],
nade_hidden_units=128,
dropout_keep_prob=0.5,
clip_norm=5,
learning_rate=0.001)),
'rnn-nade_attn': events_rnn_model.EventSequenceRnnConfig(
generator_pb2.GeneratorDetails(
id='rnn-nade_attn',
description='RNN-NADE with attention.'),
mm.PianorollEncoderDecoder(),
contrib_training.HParams(
branch_factor,
steps_per_iteration,
modify_events_callback=modify_events_callback)
def polyphonic_sequence_log_likelihood(self, sequence):
"""Evaluate the log likelihood of a polyphonic sequence.
Args:
sequence: The PolyphonicSequence object for which to evaluate the log
likelihood.
Returns:
The log likelihood of `sequence` under this model.
"""
return self._evaluate_log_likelihood([sequence])[0]
default_configs = {
'polyphony':
events_rnn_model.EventSequenceRnnConfig(
generator_pb2.GeneratorDetails(id='polyphony',
description='Polyphonic RNN'),
magenta.music.OneHotEventSequenceEncoderDecoder(
polyphony_encoder_decoder.PolyphonyOneHotEncoding()),
contrib_training.HParams(batch_size=64,
rnn_layer_sizes=[256, 256, 256],
dropout_keep_prob=0.5,
clip_norm=5,
learning_rate=0.001)),
}
