import time
"""
###################################################################
# Configuration #
###################################################################
"""
# default configuraction. Based on Magenta
default_configs = {
'drum_kit':
events_rnn_model.EventSequenceRnnConfig(
magenta.music.protobuf.generator_pb2.GeneratorDetails(
id='drum_kit',
description='Drums RNN with multiple drums and binary counters.'),
magenta.music.LookbackEventSequenceEncoderDecoder(
magenta.music.MultiDrumOneHotEncoding(),
lookback_distances=[],
binary_counter_bits=6),
contrib_training.HParams(batch_size=64,
rnn_layer_sizes=[256, 256],
dropout_keep_prob=0.5,
attn_length=32,
clip_norm=3,
learning_rate=0.001))
}
# TensorFlow Flags.
#Flags to configure evaluation and training. Spanish edition
Flags = tf.app.flags.FLAGS
tf.app.flags.DEFINE_string(
'run_dir', '/tmp/Deepbeat/logdir/run1',
'Directorio donde se encuentran los checkpoinst y '
'los análisis de los eventos.Se crearán en directorios separados para '
The log likelihood of `drums` under this model.
"""
return self._evaluate_log_likelihood([drums])[0]
# Default configurations.
default_configs = {
'one_drum':
events_rnn_model.EventSequenceRnnConfig(
magenta.protobuf.generator_pb2.GeneratorDetails(
id='one_drum', description='Drums RNN with 2-state encoding.'),
magenta.music.OneHotEventSequenceEncoderDecoder(
magenta.music.MultiDrumOneHotEncoding([
[39] + # use hand clap as default when decoding
range(mm.MIN_MIDI_PITCH, 39) + range(39, mm.MAX_MIDI_PITCH + 1)
])),
magenta.common.HParams(batch_size=128,
rnn_layer_sizes=[128, 128],
dropout_keep_prob=0.5,
skip_first_n_losses=0,
clip_norm=5,
initial_learning_rate=0.001,
decay_steps=1000,
decay_rate=0.95)),
'drum_kit':
events_rnn_model.EventSequenceRnnConfig(
magenta.protobuf.generator_pb2.GeneratorDetails(
id='drum_kit',
description='Drums RNN with multiple drums and binary counters.'),
magenta.music.LookbackEventSequenceEncoderDecoder(
magenta.music.MultiDrumOneHotEncoding(),
lookback_distances=[],
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(
magenta.protobuf.generator_pb2.GeneratorDetails(
id='polyphony', description='Polyphonic RNN'),
magenta.music.OneHotEventSequenceEncoderDecoder(
polyphony_encoder_decoder.PolyphonyOneHotEncoding()),
tf.contrib.training.HParams(batch_size=64,
rnn_layer_sizes=[256, 256, 256],
dropout_keep_prob=0.5,
clip_norm=5,
learning_rate=0.001)),
}
The log likelihood of `drums` under this model.
"""
return self._evaluate_log_likelihood([drums])[0]
# Default configurations.
default_configs = {
'one_drum':
events_rnn_model.EventSequenceRnnConfig(
magenta.music.protobuf.generator_pb2.GeneratorDetails(
id='one_drum', description='Drums RNN with 2-state encoding.'),
magenta.music.OneHotEventSequenceEncoderDecoder(
magenta.music.MultiDrumOneHotEncoding(
[[39] + # use hand clap as default when decoding
list(range(mm.MIN_MIDI_PITCH, 39)) +
list(range(39, mm.MAX_MIDI_PITCH + 1))])),
contrib_training.HParams(
batch_size=128,
rnn_layer_sizes=[128, 128],
dropout_keep_prob=0.5,
clip_norm=5,
learning_rate=0.001),
steps_per_quarter=2),
'drum_kit':
events_rnn_model.EventSequenceRnnConfig(
magenta.music.protobuf.generator_pb2.GeneratorDetails(
id='drum_kit',
description='Drums RNN with multiple drums and binary counters.'
),
magenta.music.LookbackEventSequenceEncoderDecoder(
magenta.music.MultiDrumOneHotEncoding(),
def setUp(self):
self.config = events_rnn_model.EventSequenceRnnConfig(
None, mm.PianorollEncoderDecoder(88),
tf.contrib.training.HParams())
def setUp(self):
super(PianorollPipelineTest, self).setUp()
self.config = events_rnn_model.EventSequenceRnnConfig(
None, note_seq.PianorollEncoderDecoder(88), contrib_training.HParams())
def setUp(self):
self.config = events_rnn_model.EventSequenceRnnConfig(
None,
magenta.music.OneHotEventSequenceEncoderDecoder(
polyphony_encoder_decoder.PolyphonyOneHotEncoding()),
magenta.common.HParams())
def setUp(self):
self.config = events_rnn_model.EventSequenceRnnConfig(
None,
magenta.music.OneHotEventSequenceEncoderDecoder(
magenta.music.MultiDrumOneHotEncoding()),
tf.contrib.training.HParams())
"""
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(
magenta.protobuf.generator_pb2.GeneratorDetails(
id='rnn-nade', description='RNN-NADE'),
mm.PianorollEncoderDecoder(),
tf.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(
magenta.protobuf.generator_pb2.GeneratorDetails(
id='rnn-nade_attn', description='RNN-NADE with attention.'),
mm.PianorollEncoderDecoder(),
tf.contrib.training.HParams(batch_size=48,
rnn_layer_sizes=[128, 128],
attn_length=32,
nade_hidden_units=128,
dropout_keep_prob=0.5,
clip_norm=5,
def setUp(self):
self.config = events_rnn_model.EventSequenceRnnConfig(
None, mm.PianorollEncoderDecoder(88), magenta.common.HParams())
beam_size: An integer, beam size to use when generating tracks via
beam search.
branch_factor: An integer, beam search branch factor to use.
steps_per_iteration: An integer, number of steps to take per beam search
iteration.
Returns:
The generated PolyphonicSequence object (which begins with the provided
primer track).
"""
return self._generate_events(num_steps, primer_sequence, temperature,
beam_size, branch_factor,
steps_per_iteration)
default_configs = {
'polyphony':
events_rnn_model.EventSequenceRnnConfig(
magenta.protobuf.generator_pb2.GeneratorDetails(
id='polyphony', description='Polyphonic RNN'),
magenta.music.OneHotEventSequenceEncoderDecoder(
polyphony_encoder_decoder.PolyphonyOneHotEncoding()),
magenta.common.HParams(batch_size=64,
rnn_layer_sizes=[256, 256, 256],
dropout_keep_prob=0.5,
skip_first_n_losses=10,
clip_norm=5,
initial_learning_rate=0.001,
decay_steps=1000,
decay_rate=0.95)),
}
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(
magenta.protobuf.generator_pb2.GeneratorDetails(
id='rnn-nade', description='RNN-NADE'),
mm.PianorollEncoderDecoder(),
magenta.common.HParams(batch_size=64,
rnn_layer_sizes=[128, 128, 128],
nade_hidden_units=128,
dropout_keep_prob=0.5,
skip_first_n_losses=10,
clip_norm=5,
initial_learning_rate=0.001,
decay_steps=1000,
decay_rate=0.95)),
'rnn-nade_attn':
events_rnn_model.EventSequenceRnnConfig(
magenta.protobuf.generator_pb2.GeneratorDetails(
id='rnn-nade_attn', description='RNN-NADE with attention.'),
mm.PianorollEncoderDecoder(),
magenta.common.HParams(batch_size=48,
rnn_layer_sizes=[128, 128],
attn_length=32,
nade_hidden_units=128,
def setUp(self):
super().setUp()
self.config = events_rnn_model.EventSequenceRnnConfig(
None,
note_seq.OneHotEventSequenceEncoderDecoder(
note_seq.MultiDrumOneHotEncoding()), contrib_training.HParams())