def _generate(self, input_sequence, generator_options):
if len(generator_options.input_sections) > 1:
raise sequence_generator.SequenceGeneratorError(
'This model supports at most one input_sections message, but got %s'
% len(generator_options.input_sections))
if len(generator_options.generate_sections) != 1:
raise sequence_generator.SequenceGeneratorError(
'This model supports only 1 generate_sections message, but got %s'
% len(generator_options.generate_sections))
generate_section = generator_options.generate_sections[0]
if generator_options.input_sections:
input_section = generator_options.input_sections[0]
primer_sequence = note_seq.trim_note_sequence(
input_sequence, input_section.start_time,
input_section.end_time)
input_start_step = note_seq.quantize_to_step(
input_section.start_time,
self.steps_per_second,
quantize_cutoff=0.0)
else:
primer_sequence = input_sequence
input_start_step = 0
if primer_sequence.notes:
last_end_time = max(n.end_time for n in primer_sequence.notes)
else:
last_end_time = 0
if last_end_time > generate_section.start_time:
raise sequence_generator.SequenceGeneratorError(
'Got GenerateSection request for section that is before or equal to '
'the end of the NoteSequence. This model can only extend sequences. '
'Requested start time: %s, Final note end time: %s' %
(generate_section.start_time, last_end_time))
# Quantize the priming sequence.
quantized_primer_sequence = note_seq.quantize_note_sequence_absolute(
primer_sequence, self.steps_per_second)
extracted_perfs, _ = performance_pipeline.extract_performances(
quantized_primer_sequence,
start_step=input_start_step,
num_velocity_bins=self.num_velocity_bins,
note_performance=self._note_performance)
assert len(extracted_perfs) <= 1
generate_start_step = note_seq.quantize_to_step(
generate_section.start_time,
self.steps_per_second,
quantize_cutoff=0.0)
# Note that when quantizing end_step, we set quantize_cutoff to 1.0 so it
# always rounds down. This avoids generating a sequence that ends at 5.0
# seconds when the requested end time is 4.99.
generate_end_step = note_seq.quantize_to_step(
generate_section.end_time,
self.steps_per_second,
quantize_cutoff=1.0)
if extracted_perfs and extracted_perfs[0]:
performance = extracted_perfs[0]
else:
# If no track could be extracted, create an empty track that starts at the
# requested generate_start_step.
performance = note_seq.Performance(
steps_per_second=(quantized_primer_sequence.quantization_info.
steps_per_second),
start_step=generate_start_step,
num_velocity_bins=self.num_velocity_bins)
# Ensure that the track extends up to the step we want to start generating.
performance.set_length(generate_start_step - performance.start_step)
# Extract generation arguments from generator options.
arg_types = {
'disable_conditioning':
lambda arg: ast.literal_eval(arg.string_value),
'temperature': lambda arg: arg.float_value,
'beam_size': lambda arg: arg.int_value,
'branch_factor': lambda arg: arg.int_value,
'steps_per_iteration': lambda arg: arg.int_value
}
if self.control_signals:
for control in self.control_signals:
arg_types[control.name] = lambda arg: ast.literal_eval(
arg.string_value)
args = dict((name, value_fn(generator_options.args[name]))
for name, value_fn in arg_types.items()
if name in generator_options.args)
# Make sure control signals are present and convert to lists if necessary.
if self.control_signals:
for control in self.control_signals:
if control.name not in args:
tf.logging.warning(
'Control value not specified, using default: %s = %s',
control.name, control.default_value)
args[control.name] = [control.default_value]
elif control.validate(args[control.name]):
args[control.name] = [args[control.name]]
else:
if not isinstance(args[control.name], list) or not all(
control.validate(value)
for value in args[control.name]):
tf.logging.fatal('Invalid control value: %s = %s',
control.name, args[control.name])
# Make sure disable conditioning flag is present when conditioning is
# optional and convert to list if necessary.
if self.optional_conditioning:
if 'disable_conditioning' not in args:
args['disable_conditioning'] = [False]
elif isinstance(args['disable_conditioning'], bool):
args['disable_conditioning'] = [args['disable_conditioning']]
else:
if not isinstance(
args['disable_conditioning'], list) or not all(
isinstance(value, bool)
for value in args['disable_conditioning']):
tf.logging.fatal('Invalid disable_conditioning value: %s',
args['disable_conditioning'])
total_steps = performance.num_steps + (generate_end_step -
generate_start_step)
if 'notes_per_second' in args:
mean_note_density = (sum(args['notes_per_second']) /
len(args['notes_per_second']))
else:
mean_note_density = DEFAULT_NOTE_DENSITY
# Set up functions that map generation step to control signal values and
# disable conditioning flag.
if self.control_signals:
control_signal_fns = []
for control in self.control_signals:
control_signal_fns.append(
functools.partial(_step_to_value,
num_steps=total_steps,
values=args[control.name]))
del args[control.name]
args['control_signal_fns'] = control_signal_fns
if self.optional_conditioning:
args['disable_conditioning_fn'] = functools.partial(
_step_to_value,
num_steps=total_steps,
values=args['disable_conditioning'])
del args['disable_conditioning']
if not performance:
# Primer is empty; let's just start with silence.
performance.set_length(
min(performance.max_shift_steps, total_steps))
while performance.num_steps < total_steps:
# Assume the average specified (or default) note density and 4 RNN steps
# per note. Can't know for sure until generation is finished because the
# number of notes per quantized step is variable.
note_density = max(1.0, mean_note_density)
steps_to_gen = total_steps - performance.num_steps
rnn_steps_to_gen = int(
math.ceil(4.0 * note_density * steps_to_gen /
self.steps_per_second))
tf.logging.info(
'Need to generate %d more steps for this sequence, will try asking '
'for %d RNN steps' % (steps_to_gen, rnn_steps_to_gen))
performance = self._model.generate_performance(
len(performance) + rnn_steps_to_gen, performance, **args)
if not self.fill_generate_section:
# In the interest of speed just go through this loop once, which may not
# entirely fill the generate section.
break
performance.set_length(total_steps)
generated_sequence = performance.to_sequence(
max_note_duration=self.max_note_duration)
assert (generated_sequence.total_time -
generate_section.end_time) <= 1e-5
return generated_sequence
def _generate(self, input_sequence, generator_options):
if len(generator_options.input_sections) > 1:
raise sequence_generator.SequenceGeneratorError(
'This model supports at most one input_sections message, but got %s'
% len(generator_options.input_sections))
if len(generator_options.generate_sections) != 1:
raise sequence_generator.SequenceGeneratorError(
'This model supports only 1 generate_sections message, but got %s'
% len(generator_options.generate_sections))
# This sequence will be quantized later, so it is guaranteed to have only 1
# tempo.
qpm = note_seq.DEFAULT_QUARTERS_PER_MINUTE
if input_sequence.tempos:
qpm = input_sequence.tempos[0].qpm
steps_per_second = note_seq.steps_per_quarter_to_steps_per_second(
self.steps_per_quarter, qpm)
generate_section = generator_options.generate_sections[0]
if generator_options.input_sections:
input_section = generator_options.input_sections[0]
primer_sequence = note_seq.trim_note_sequence(
input_sequence, input_section.start_time,
input_section.end_time)
input_start_step = note_seq.quantize_to_step(
input_section.start_time, steps_per_second, quantize_cutoff=0)
else:
primer_sequence = input_sequence
input_start_step = 0
if primer_sequence.notes:
last_end_time = max(n.end_time for n in primer_sequence.notes)
else:
last_end_time = 0
if last_end_time > generate_section.start_time:
raise sequence_generator.SequenceGeneratorError(
'Got GenerateSection request for section that is before or equal to '
'the end of the NoteSequence. This model can only extend sequences. '
'Requested start time: %s, Final note end time: %s' %
(generate_section.start_time, last_end_time))
# Quantize the priming sequence.
quantized_primer_sequence = note_seq.quantize_note_sequence(
primer_sequence, self.steps_per_quarter)
extracted_seqs, _ = polyphony_lib.extract_polyphonic_sequences(
quantized_primer_sequence, start_step=input_start_step)
assert len(extracted_seqs) <= 1
generate_start_step = note_seq.quantize_to_step(
generate_section.start_time, steps_per_second, quantize_cutoff=0)
# Note that when quantizing end_step, we set quantize_cutoff to 1.0 so it
# always rounds down. This avoids generating a sequence that ends at 5.0
# seconds when the requested end time is 4.99.
generate_end_step = note_seq.quantize_to_step(
generate_section.end_time, steps_per_second, quantize_cutoff=1.0)
if extracted_seqs and extracted_seqs[0]:
poly_seq = extracted_seqs[0]
else:
# If no track could be extracted, create an empty track that starts at the
# requested generate_start_step. This will result in a sequence that
# contains only the START token.
poly_seq = polyphony_lib.PolyphonicSequence(
steps_per_quarter=(quantized_primer_sequence.quantization_info.
steps_per_quarter),
start_step=generate_start_step)
# Ensure that the track extends up to the step we want to start generating.
poly_seq.set_length(generate_start_step - poly_seq.start_step)
# Trim any trailing end events to prepare the sequence for more events to be
# appended during generation.
poly_seq.trim_trailing_end_events()
# Extract generation arguments from generator options.
arg_types = {
'temperature': lambda arg: arg.float_value,
'beam_size': lambda arg: arg.int_value,
'branch_factor': lambda arg: arg.int_value,
'steps_per_iteration': lambda arg: arg.int_value
}
args = dict((name, value_fn(generator_options.args[name]))
for name, value_fn in arg_types.items()
if name in generator_options.args)
# Inject the priming sequence as melody in the output of the generator, if
# requested.
# This option starts with no_ so that if it is unspecified (as will be the
# case when used with the midi interface), the default will be to inject the
# primer.
if not (generator_options.args['no_inject_primer_during_generation'].
bool_value):
melody_to_inject = copy.deepcopy(poly_seq)
if generator_options.args['condition_on_primer'].bool_value:
inject_start_step = poly_seq.num_steps
else:
# 0 steps because we'll overwrite poly_seq with a blank sequence below.
inject_start_step = 0
args['modify_events_callback'] = functools.partial(
_inject_melody, melody_to_inject, inject_start_step)
# If we don't want to condition on the priming sequence, then overwrite
# poly_seq with a blank sequence to feed into the generator.
if not generator_options.args['condition_on_primer'].bool_value:
poly_seq = polyphony_lib.PolyphonicSequence(
steps_per_quarter=(quantized_primer_sequence.quantization_info.
steps_per_quarter),
start_step=generate_start_step)
poly_seq.trim_trailing_end_events()
total_steps = poly_seq.num_steps + (generate_end_step -
generate_start_step)
while poly_seq.num_steps < total_steps:
# Assume it takes ~5 rnn steps to generate one quantized step.
# Can't know for sure until generation is finished because the number of
# notes per quantized step is variable.
steps_to_gen = total_steps - poly_seq.num_steps
rnn_steps_to_gen = 5 * steps_to_gen
tf.logging.info(
'Need to generate %d more steps for this sequence, will try asking '
'for %d RNN steps' % (steps_to_gen, rnn_steps_to_gen))
poly_seq = self._model.generate_polyphonic_sequence(
len(poly_seq) + rnn_steps_to_gen, poly_seq, **args)
poly_seq.set_length(total_steps)
if generator_options.args['condition_on_primer'].bool_value:
generated_sequence = poly_seq.to_sequence(qpm=qpm)
else:
# Specify a base_note_sequence because the priming sequence was not
# included in poly_seq.
generated_sequence = poly_seq.to_sequence(
qpm=qpm, base_note_sequence=copy.deepcopy(primer_sequence))
assert (generated_sequence.total_time -
generate_section.end_time) <= 1e-5
return generated_sequence
def _generate(self, input_sequence, generator_options):
if len(generator_options.input_sections) > 1:
raise sequence_generator.SequenceGeneratorError(
'This model supports at most one input_sections message, but got %s' %
len(generator_options.input_sections))
if len(generator_options.generate_sections) != 1:
raise sequence_generator.SequenceGeneratorError(
'This model supports only 1 generate_sections message, but got %s' %
len(generator_options.generate_sections))
if input_sequence and input_sequence.tempos:
qpm = input_sequence.tempos[0].qpm
else:
qpm = note_seq.DEFAULT_QUARTERS_PER_MINUTE
steps_per_second = note_seq.steps_per_quarter_to_steps_per_second(
self.steps_per_quarter, qpm)
generate_section = generator_options.generate_sections[0]
if generator_options.input_sections:
input_section = generator_options.input_sections[0]
primer_sequence = note_seq.trim_note_sequence(input_sequence,
input_section.start_time,
input_section.end_time)
input_start_step = note_seq.quantize_to_step(
input_section.start_time, steps_per_second, quantize_cutoff=0)
else:
primer_sequence = input_sequence
input_start_step = 0
if primer_sequence.notes:
last_end_time = max(n.end_time for n in primer_sequence.notes)
else:
last_end_time = 0
if last_end_time > generate_section.start_time:
raise sequence_generator.SequenceGeneratorError(
'Got GenerateSection request for section that is before the end of '
'the NoteSequence. This model can only extend sequences. Requested '
'start time: %s, Final note end time: %s' %
(generate_section.start_time, last_end_time))
# Quantize the priming sequence.
quantized_sequence = note_seq.quantize_note_sequence(
primer_sequence, self.steps_per_quarter)
# Setting gap_bars to infinite ensures that the entire input will be used.
extracted_melodies, _ = melody_pipelines.extract_melodies(
quantized_sequence, search_start_step=input_start_step, min_bars=0,
min_unique_pitches=1, gap_bars=float('inf'),
ignore_polyphonic_notes=True)
assert len(extracted_melodies) <= 1
start_step = note_seq.quantize_to_step(
generate_section.start_time, steps_per_second, quantize_cutoff=0)
# Note that when quantizing end_step, we set quantize_cutoff to 1.0 so it
# always rounds down. This avoids generating a sequence that ends at 5.0
# seconds when the requested end time is 4.99.
end_step = note_seq.quantize_to_step(
generate_section.end_time, steps_per_second, quantize_cutoff=1.0)
if extracted_melodies and extracted_melodies[0]:
melody = extracted_melodies[0]
else:
# If no melody could be extracted, create an empty melody that starts 1
# step before the request start_step. This will result in 1 step of
# silence when the melody is extended below.
steps_per_bar = int(
note_seq.steps_per_bar_in_quantized_sequence(quantized_sequence))
melody = note_seq.Melody([],
start_step=max(0, start_step - 1),
steps_per_bar=steps_per_bar,
steps_per_quarter=self.steps_per_quarter)
# Ensure that the melody extends up to the step we want to start generating.
melody.set_length(start_step - melody.start_step)
# Extract generation arguments from generator options.
arg_types = {
'temperature': lambda arg: arg.float_value,
'beam_size': lambda arg: arg.int_value,
'branch_factor': lambda arg: arg.int_value,
'steps_per_iteration': lambda arg: arg.int_value
}
args = dict((name, value_fn(generator_options.args[name]))
for name, value_fn in arg_types.items()
if name in generator_options.args)
generated_melody = self._model.generate_melody(
end_step - melody.start_step, melody, **args)
generated_sequence = generated_melody.to_sequence(qpm=qpm)
assert (generated_sequence.total_time - generate_section.end_time) <= 1e-5
return generated_sequence
def _generate(self, input_sequence, generator_options):
if len(generator_options.input_sections) > 1:
raise sequence_generator.SequenceGeneratorError(
'This model supports at most one input_sections message, but got %s'
% len(generator_options.input_sections))
if len(generator_options.generate_sections) != 1:
raise sequence_generator.SequenceGeneratorError(
'This model supports only 1 generate_sections message, but got %s'
% len(generator_options.generate_sections))
# This sequence will be quantized later, so it is guaranteed to have only 1
# tempo.
qpm = note_seq.DEFAULT_QUARTERS_PER_MINUTE
if input_sequence.tempos:
qpm = input_sequence.tempos[0].qpm
steps_per_second = note_seq.steps_per_quarter_to_steps_per_second(
self.steps_per_quarter, qpm)
generate_section = generator_options.generate_sections[0]
if generator_options.input_sections:
input_section = generator_options.input_sections[0]
primer_sequence = note_seq.trim_note_sequence(
input_sequence, input_section.start_time,
input_section.end_time)
input_start_step = note_seq.quantize_to_step(
input_section.start_time, steps_per_second, quantize_cutoff=0)
else:
primer_sequence = input_sequence
input_start_step = 0
if primer_sequence.notes:
last_end_time = max(n.end_time for n in primer_sequence.notes)
else:
last_end_time = 0
if last_end_time > generate_section.start_time:
raise sequence_generator.SequenceGeneratorError(
'Got GenerateSection request for section that is before or equal to '
'the end of the NoteSequence. This model can only extend sequences. '
'Requested start time: %s, Final note end time: %s' %
(generate_section.start_time, last_end_time))
# Quantize the priming sequence.
quantized_primer_sequence = note_seq.quantize_note_sequence(
primer_sequence, self.steps_per_quarter)
extracted_seqs, _ = pianoroll_pipeline.extract_pianoroll_sequences(
quantized_primer_sequence, start_step=input_start_step)
assert len(extracted_seqs) <= 1
generate_start_step = note_seq.quantize_to_step(
generate_section.start_time, steps_per_second, quantize_cutoff=0)
# Note that when quantizing end_step, we set quantize_cutoff to 1.0 so it
# always rounds down. This avoids generating a sequence that ends at 5.0
# seconds when the requested end time is 4.99.
generate_end_step = note_seq.quantize_to_step(
generate_section.end_time, steps_per_second, quantize_cutoff=1.0)
if extracted_seqs and extracted_seqs[0]:
pianoroll_seq = extracted_seqs[0]
else:
raise ValueError('No priming pianoroll could be extracted.')
# Ensure that the track extends up to the step we want to start generating.
pianoroll_seq.set_length(generate_start_step -
pianoroll_seq.start_step)
# Extract generation arguments from generator options.
arg_types = {
'beam_size': lambda arg: arg.int_value,
'branch_factor': lambda arg: arg.int_value,
}
args = dict((name, value_fn(generator_options.args[name]))
for name, value_fn in arg_types.items()
if name in generator_options.args)
total_steps = pianoroll_seq.num_steps + (generate_end_step -
generate_start_step)
pianoroll_seq = self._model.generate_pianoroll_sequence(
total_steps, pianoroll_seq, **args)
pianoroll_seq.set_length(total_steps)
generated_sequence = pianoroll_seq.to_sequence(qpm=qpm)
assert (generated_sequence.total_time -
generate_section.end_time) <= 1e-5
return generated_sequence