def get_primer_ns(filename, max_length):
"""
Convert Midi file to note sequences for priming.
:param filename: Midi file name.
:param max_length: Maximum note sequence length for priming in seconds.
:return:
Note sequences for priming.
"""
primer_ns = mm.midi_file_to_note_sequence(filename)
# Handle sustain pedal in primer.
primer_ns = mm.apply_sustain_control_changes(primer_ns)
# Trim to desired number of seconds.
if primer_ns.total_time > max_length:
LOGGER.warn(
'Primer duration %d is longer than max second %d, truncating.' %
(primer_ns.total_time, max_length))
primer_ns = mm.extract_subsequence(primer_ns, 0, max_length)
# Remove drums from primer if present.
if any(note.is_drum for note in primer_ns.notes):
LOGGER.warn('Primer contains drums; they will be removed.')
notes = [note for note in primer_ns.notes if not note.is_drum]
del primer_ns.notes[:]
primer_ns.notes.extend(notes)
# Set primer instrument and program.
for note in primer_ns.notes:
note.instrument = 1
note.program = 0
return primer_ns
def _read_midi(self, input_file):
"""
Read midi file into note sequence.
:param input_file: A string path to midi file.
:return: Note sequence.
"""
note_sequence = mm.midi_file_to_note_sequence(input_file)
# Handle sustain pedal in primer.
if self.config.sustain:
note_sequence = mm.apply_sustain_control_changes(note_sequence)
# Trim to desired number of seconds.
if note_sequence.total_time > self.config.max_length:
LOGGER.warning(
'Note sequence %d is longer than max seconds %d, truncating.',
note_sequence.total_time, self.config.max_length)
note_sequence = mm.extract_subsequence(note_sequence, 0,
self.config.max_length)
# Whether or not remove drums.
if any(note.is_drum
for note in note_sequence.notes) and not self.config.use_drum:
LOGGER.warning('Midi file contains drum sounds, removing.')
notes = [note for note in note_sequence.notes if not note.is_drum]
del note_sequence.notes[:]
note_sequence.notes.extend(notes)
# Set primer instrument and program.
for note in note_sequence.notes:
note.instrument = 1
note.program = 0
return note_sequence
def piano_continuation(primer):
print("이함수도안돌아?")
primer_ns = mm.midi_file_to_note_sequence(primer)
# Handle sustain pedal in the primer.
primer_ns = mm.apply_sustain_control_changes(primer_ns)
# Trim to desired number of seconds.
max_primer_seconds = 20 #@param {type:"slider", min:1, max:120}
if primer_ns.total_time > max_primer_seconds:
print('Primer is longer than %d seconds, truncating.' %
max_primer_seconds)
primer_ns = mm.extract_subsequence(primer_ns, 0, max_primer_seconds)
# Remove drums from primer if present.
if any(note.is_drum for note in primer_ns.notes):
print('Primer contains drums; they will be removed.')
notes = [note for note in primer_ns.notes if not note.is_drum]
del primer_ns.notes[:]
primer_ns.notes.extend(notes)
# Set primer instrument and program.
for note in primer_ns.notes:
note.instrument = 1
note.program = 0
#note_sequence_to_midi_file(primer_ns, 'modified_'+primer)
targets = uncondi_encoders['targets'].encode_note_sequence(primer_ns)
# Remove the end token from the encoded primer.
targets = targets[:-1]
decode_length = max(0, 4096 - len(targets))
if len(targets) >= 4096:
print(
'Primer has more events than maximum sequence length; nothing will be generated.'
)
# Generate sample events.
sample_ids = next(uncondi_samples)['outputs']
# Decode to NoteSequence.
midi_filename = decode(sample_ids, encoder=uncondi_encoders['targets'])
ns = mm.midi_file_to_note_sequence(midi_filename)
# Append continuation to primer.
continuation_ns = mm.concatenate_sequences([primer_ns, ns])
note_sequence_to_midi_file(continuation_ns, 'continuated_' + primer)
def generate_primer(self):
"""
Put something important here.
"""
if self.conditioned:
raise ValueError("Should be using an unconditioned model!")
primer_ns = self.sequence
primer_ns = mm.apply_sustain_control_changes(primer_ns)
max_primer_seconds = 10
if primer_ns.total_time > max_primer_seconds:
print(f'Primer is longer than {max_primer_seconds} seconds, truncating.')
# cut primer if it's too long
primer_ns = mm.extract_subsequence(
primer_ns, 0, max_primer_seconds)
if any(note.is_drum for note in primer_ns.notes):
print('Primer contains drums; they will be removed.')
notes = [note for note in primer_ns.notes if not note.is_drum]
del primer_ns.notes[:]
primer_ns.notes.extend(notes)
for note in primer_ns.notes:
# make into piano
note.instrument = 1
note.program = 0
self.targets = self.encoders['targets'].encode_note_sequence(
primer_ns)
# Remove the end token from the encoded primer.
self.targets = self.targets[:-1]
self.decode_length = max(0, 4096 - len(self.targets))
if len(self.targets) >= 4096:
print('Primer has more events than maximum sequence length; nothing will be generated.')
# Generate sample events.
sample_ids = next(self.samples)['outputs']
midi_filename = self.decode(
sample_ids,
encoder=self.encoders['targets'])
ns = mm.midi_file_to_note_sequence(midi_filename)
# Append continuation to primer.
continuation_ns = mm.concatenate_sequences([primer_ns, ns])
request_dict = self.put_request_dict
generated_sequence_2_mp3(continuation_ns, f"{self.unique_id}", use_salamander=True,
request_dict=request_dict)
def generate_basic_notes(self, qpm=160, failsafe=False):
"""
Requires melody conditioned model.
"""
if not self.conditioned:
raise ValueError("Model should be conditioned!")
if failsafe:
self.failsafe()
else:
melody_ns = copy.deepcopy(self.sequence)
try:
melody_instrument = mm.infer_melody_for_sequence(melody_ns)
notes = [note for note in melody_ns.notes
if note.instrument == melody_instrument]
melody_ns.notes.extend(
sorted(notes, key=lambda note: note.start_time))
for i in range(len(melody_ns.notes) - 1):
melody_ns.notes[i].end_time = melody_ns.notes[i + 1].start_time
# sequence can only be one min to save time during inference.
melody_ns = mm.extract_subsequence(melody_ns, 0, 60)
self.inputs = self.encoders['inputs'].encode_note_sequence(
melody_ns)
print("Melody successfully parsed and encoded!")
except Exception as e:
print(f"Error in encoding stage {e}")
print("Resorting to a basic melody")
self.failsafe()
self.decode_length = 4096
sample_ids = next(self.samples)['outputs']
# Decode to NoteSequence.
midi_filename = self.decode(
sample_ids,
encoder=self.encoders['targets'])
accompaniment_ns = mm.midi_file_to_note_sequence(midi_filename)
request_dict = self.put_request_dict
generated_sequence_2_mp3(accompaniment_ns, f"{self.unique_id}", use_salamander=True,
request_dict=request_dict)
def _generate(self, input_sequence, generator_options):
if len(generator_options.input_sections) > 1:
raise mm.SequenceGeneratorException(
'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 mm.SequenceGeneratorException(
'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 = mm.DEFAULT_QUARTERS_PER_MINUTE
if input_sequence.tempos:
qpm = input_sequence.tempos[0].qpm
generate_section = generator_options.generate_sections[0]
if generator_options.input_sections:
input_section = generator_options.input_sections[0]
primer_sequence = mm.extract_subsequence(input_sequence,
input_section.start_time,
input_section.end_time)
input_start_step = self.seconds_to_steps(input_section.start_time,
qpm)
else:
primer_sequence = input_sequence
input_start_step = 0
last_end_time = (max(
n.end_time
for n in primer_sequence.notes) if primer_sequence.notes else 0)
if last_end_time > generate_section.start_time:
raise mm.SequenceGeneratorException(
'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 = mm.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 = self.seconds_to_steps(
generate_section.start_time, qpm)
generate_end_step = self.seconds_to_steps(generate_section.end_time,
qpm)
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.
# Note that start_step is 0 because we overwrite poly_seq below. If we
# included the priming sequence in poly_seq, it would be poly_seq.num_steps.
melody_to_inject = copy.deepcopy(poly_seq)
args['modify_events_callback'] = partial(_inject_melody,
melody_to_inject, 0)
# Overwrite poly_seq with a blank sequence to feed into the generator so it
# is conditioned only on the melody events that are injected as the sequence
# is created. Otherwise, the generator would have to determine the most
# likely sequence to follow a monophonic line, which is something not
# present in the current training data (Bach Chorales).
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()
# If we wanted to include the priming sequence and didn't clear poly_seq
# above, this is how we would calculate total_steps.
# total_steps = poly_seq.num_steps + (
# generate_end_step - generate_start_step)
total_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)
# Specify a base_note_sequence because the priming sequence is not included
# in poly_seq. If we did not clear poly_seq above, then we would not want to
# specify a base_note_sequence.
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 mm.SequenceGeneratorException(
'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 mm.SequenceGeneratorException(
'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 = mm.DEFAULT_QUARTERS_PER_MINUTE
if input_sequence.tempos:
qpm = input_sequence.tempos[0].qpm
generate_section = generator_options.generate_sections[0]
if generator_options.input_sections:
input_section = generator_options.input_sections[0]
primer_sequence = mm.extract_subsequence(input_sequence,
input_section.start_time,
input_section.end_time)
input_start_step = self.seconds_to_steps(input_section.start_time,
qpm)
else:
primer_sequence = input_sequence
input_start_step = 0
last_end_time = (max(
n.end_time
for n in primer_sequence.notes) if primer_sequence.notes else 0)
if last_end_time > generate_section.start_time:
raise mm.SequenceGeneratorException(
'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 = mm.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
start_step = self.seconds_to_steps(generate_section.start_time, qpm)
end_step = self.seconds_to_steps(generate_section.end_time, qpm)
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 1 step
# before the request start_step. This will result in 1 step of silence
# when the track is extended below.
poly_seq = polyphony_lib.PolyphonicSequence(steps_per_quarter=(
quantized_primer_sequence.quantization_info.steps_per_quarter),
start_step=start_step)
# Ensure that the track extends up to the step we want to start generating.
poly_seq.set_length(start_step - poly_seq.start_step)
poly_seq.trim_trailing_end_and_step_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)
total_steps = end_step - 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)
generated_sequence = poly_seq.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 mm.SequenceGeneratorException(
'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 mm.SequenceGeneratorException(
'This model supports only 1 generate_sections message, but got %s' %
len(generator_options.generate_sections))
qpm = (input_sequence.tempos[0].qpm
if input_sequence and input_sequence.tempos
else mm.DEFAULT_QUARTERS_PER_MINUTE)
generate_section = generator_options.generate_sections[0]
if generator_options.input_sections:
input_section = generator_options.input_sections[0]
primer_sequence = mm.extract_subsequence(
input_sequence, input_section.start_time, input_section.end_time)
input_start_step = self._seconds_to_steps(input_section.start_time, qpm)
else:
primer_sequence = input_sequence
input_start_step = 0
last_end_time = (max(n.end_time for n in primer_sequence.notes)
if primer_sequence.notes else 0)
if last_end_time >= generate_section.start_time:
raise mm.SequenceGeneratorException(
'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_sequence = mm.quantize_note_sequence(
primer_sequence, self._steps_per_quarter)
# Setting gap_bars to infinite ensures that the entire input will be used.
extracted_melodies, _ = mm.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 = self._seconds_to_steps(
generate_section.start_time, qpm)
end_step = self._seconds_to_steps(generate_section.end_time, qpm)
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.
melody = mm.Melody([], start_step=max(0, start_step - 1))
# 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 render_sequence_to_music_dict(midi_file,
music_dict,
model_string="melody_rnn"):
sequence = mm.midi_file_to_note_sequence(midi_file)
# scale to num steps.
music_dict['num_steps'] = 1024 * music_dict['length']
backup_sequence = None
basic_models = [
"melody_rnn", "performance_rnn", "polyphony_rnn", "pianoroll_rnn_nade"
]
if model_string in basic_models:
subsequence = mm.extract_subsequence(sequence, 0.0, C.SUBSEQUENCE_TIME)
for note in subsequence.notes:
# rnns can work with piano data.
note.program = 0
note.instrument = 1
music_dict['sequence'] = subsequence
if model_string == "performance_rnn":
music_dict['num_steps'] = music_dict['num_steps'] * 4
elif model_string == "improv_rnn" or model_string == "music_vae":
subsequence = mm.extract_subsequence(sequence, 0.0, C.SUBSEQUENCE_TIME)
melody = mm.infer_melody_for_sequence(subsequence)
new_sequence = music_pb2.NoteSequence()
backup_sequence = music_pb2.NoteSequence()
new_val = 0.
backup_val = 0.
for note in subsequence.notes:
# rnns can work with piano data.
if note.instrument == melody:
start = note.start_time
end = note.end_time
diff = end - start
new_sequence.notes.add(pitch=note.pitch,
start_time=new_val,
end_time=new_val + diff,
velocity=160)
backup_sequence.notes.add(pitch=note.pitch,
start_time=backup_val,
end_time=backup_val + 0.5,
velocity=160)
new_val += diff
backup_val += 0.5
if model_string == "improv_rnn":
note.program = 0
note.instrument = 1
new_sequence.total_time = new_val
new_sequence.tempos.add(qpm=subsequence.tempos[0].qpm)
backup_sequence.total_time = backup_val
backup_sequence.tempos.add(qpm=60)
music_dict['sequence'] = subsequence
music_dict['backup_sequence'] = backup_sequence
elif model_string == "music_transformer":
# model generate will take care of things
music_dict['sequence'] = sequence
return music_dict
twinkle_twinkle.notes.add(pitch=67, start_time=1.5, end_time=2.0, velocity=80)
twinkle_twinkle.notes.add(pitch=69, start_time=2.0, end_time=2.5, velocity=80)
twinkle_twinkle.notes.add(pitch=69, start_time=2.5, end_time=3.0, velocity=80)
twinkle_twinkle.notes.add(pitch=67, start_time=3.0, end_time=4.0, velocity=80)
twinkle_twinkle.notes.add(pitch=65, start_time=4.0, end_time=4.5, velocity=80)
twinkle_twinkle.notes.add(pitch=65, start_time=4.5, end_time=5.0, velocity=80)
twinkle_twinkle.notes.add(pitch=64, start_time=5.0, end_time=5.5, velocity=80)
twinkle_twinkle.notes.add(pitch=64, start_time=5.5, end_time=6.0, velocity=80)
twinkle_twinkle.notes.add(pitch=62, start_time=6.0, end_time=6.5, velocity=80)
twinkle_twinkle.notes.add(pitch=62, start_time=6.5, end_time=7.0, velocity=80)
twinkle_twinkle.notes.add(pitch=60, start_time=7.0, end_time=8.0, velocity=80)
twinkle_twinkle.total_time = 8
twinkle_twinkle.tempos.add(qpm=60)
babyshark = mm.midi_file_to_note_sequence('./mid/babyshark.mid')
babyshark = mm.extract_subsequence(babyshark, 0, 8)
babyshark.ticks_per_quarter = 0
babyshark.time_signatures.pop()
babyshark.key_signatures.pop()
babyshark.tempos.pop()
babyshark.tempos.add(qpm=60)
for note in babyshark.notes:
if note.pitch < 60:
note.pitch = 60
note.instrument = 0
note.is_drum = False
# This gives us a list of sequences.
note_sequences = music_vae.interpolate(twinkle_twinkle,
import magenta.music as mm
babyshark = mm.midi_file_to_note_sequence('./mid/babyshark_full.mid')
# truncate @ 14 seconds
babyshark = mm.extract_subsequence(babyshark, 14, 14 + 8.5)
babyshark = mm.extract_subsequence(babyshark, 0, 8)
babyshark.ticks_per_quarter = 0
babyshark.time_signatures.pop()
babyshark.key_signatures.pop()
babyshark.tempos.pop()
babyshark.tempos.add(qpm=60)
for note in babyshark.notes:
if note.pitch < 60:
note.pitch = 60
note.instrument = 0
note.is_drum = False
mm.sequence_proto_to_midi_file(babyshark, './mid/babyshark.mid')
primer = 'C major scale' #@param ['C major arpeggio', 'C major scale', 'Clair de Lune', 'Upload your own!']
if primer == 'Upload your own!':
primer_ns = upload_midi()
else:
# Use one of the provided primers.
primer_ns = mm.midi_file_to_note_sequence(filenames[primer])
# Handle sustain pedal in the primer.
primer_ns = mm.apply_sustain_control_changes(primer_ns)
# Trim to desired number of seconds.
max_primer_seconds = 20 #@param {type:"slider", min:1, max:120}
if primer_ns.total_time > max_primer_seconds:
print('Primer is longer than %d seconds, truncating.' % max_primer_seconds)
primer_ns = mm.extract_subsequence(
primer_ns, 0, max_primer_seconds)
# Remove drums from primer if present.
if any(note.is_drum for note in primer_ns.notes):
print('Primer contains drums; they will be removed.')
notes = [note for note in primer_ns.notes if not note.is_drum]
del primer_ns.notes[:]
primer_ns.notes.extend(notes)
# Set primer instrument and program.
for note in primer_ns.notes:
note.instrument = 1
note.program = 0
# Play and plot the primer.
mm.play_sequence(
def _generate(self, input_sequence, generator_options):
if len(generator_options.input_sections) > 1:
raise mm.SequenceGeneratorException(
'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 mm.SequenceGeneratorException(
'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 = mm.extract_subsequence(
input_sequence, input_section.start_time, input_section.end_time)
else:
primer_sequence = input_sequence
last_end_time = (max(n.end_time for n in primer_sequence.notes)
if primer_sequence.notes else 0)
if last_end_time >= generate_section.start_time:
raise mm.SequenceGeneratorException(
'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_sequence = mm.quantize_note_sequence(
primer_sequence, self._steps_per_quarter)
# Setting gap_bars to infinite ensures that the entire input will be used.
extracted_drum_tracks, _ = mm.extract_drum_tracks(
quantized_sequence, min_bars=0, gap_bars=float('inf'))
assert len(extracted_drum_tracks) <= 1
qpm = (primer_sequence.tempos[0].qpm
if primer_sequence and primer_sequence.tempos
else mm.DEFAULT_QUARTERS_PER_MINUTE)
start_step = self._seconds_to_steps(
generate_section.start_time, qpm)
end_step = self._seconds_to_steps(generate_section.end_time, qpm)
if extracted_drum_tracks and extracted_drum_tracks[0]:
drums = extracted_drum_tracks[0]
else:
# If no drum track could be extracted, create an empty drum track that
# starts 1 step before the request start_step. This will result in 1 step
# of silence when the drum track is extended below.
drums = mm.DrumTrack([], start_step=max(0, start_step - 1))
# Ensure that the drum track extends up to the step we want to start
# generating.
drums.set_length(start_step - drums.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_drums = self._model.generate_drum_track(
end_step - drums.start_step, drums, **args)
generated_sequence = generated_drums.to_sequence(qpm=qpm)
assert (generated_sequence.total_time - generate_section.end_time) <= 1e-5
return generated_sequence
def trim_sequences(seqs, num_seconds=BAR_SECONDS):
for i in range(len(seqs)):
seqs[i] = mm.extract_subsequence(seqs[i], 0.0, num_seconds)
seqs[i].total_time = num_seconds
def _generate(self, input_sequence, generator_options):
if len(generator_options.input_sections) > 1:
raise mm.SequenceGeneratorException(
'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 mm.SequenceGeneratorException(
'This model supports only 1 generate_sections message, but got %s' %
len(generator_options.generate_sections))
qpm = (input_sequence.tempos[0].qpm
if input_sequence and input_sequence.tempos
else mm.DEFAULT_QUARTERS_PER_MINUTE)
generate_section = generator_options.generate_sections[0]
if generator_options.input_sections:
# Use primer melody from input section only. Take backing chords from
# beginning of input section through end of generate section.
input_section = generator_options.input_sections[0]
primer_sequence = mm.extract_subsequence(
input_sequence, input_section.start_time, input_section.end_time)
backing_sequence = mm.extract_subsequence(
input_sequence, input_section.start_time, generate_section.end_time)
input_start_step = self.seconds_to_steps(input_section.start_time, qpm)
else:
# No input section. Take primer melody from the beginning of the sequence
# up until the start of the generate section.
primer_sequence = mm.extract_subsequence(
input_sequence, 0.0, generate_section.start_time)
backing_sequence = mm.extract_subsequence(
input_sequence, 0.0, generate_section.end_time)
input_start_step = 0
last_end_time = (max(n.end_time for n in primer_sequence.notes)
if primer_sequence.notes else 0)
if last_end_time >= generate_section.start_time:
raise mm.SequenceGeneratorException(
'Got GenerateSection request for section that is before or equal to '
'the end of the input section. This model can only extend melodies. '
'Requested start time: %s, Final note end time: %s' %
(generate_section.start_time, last_end_time))
# Quantize the priming and backing sequences.
quantized_primer_sequence = mm.quantize_note_sequence(
primer_sequence, self._steps_per_quarter)
quantized_backing_sequence = mm.quantize_note_sequence(
backing_sequence, self._steps_per_quarter)
# Setting gap_bars to infinite ensures that the entire input will be used.
extracted_melodies, _ = mm.extract_melodies(
quantized_primer_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 = self.seconds_to_steps(
generate_section.start_time, qpm)
end_step = self.seconds_to_steps(generate_section.end_time, qpm)
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.
melody = mm.Melody([], start_step=max(0, start_step - 1))
extracted_chords, _ = mm.extract_chords(quantized_backing_sequence)
chords = extracted_chords[0]
# Make sure that chords and melody start on the same step.
if chords.start_step < melody.start_step:
chords.set_length(len(chords) - melody.start_step + chords.start_step)
assert chords.end_step == end_step
# 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(melody, chords, **args)
generated_lead_sheet = mm.LeadSheet(generated_melody, chords)
generated_sequence = generated_lead_sheet.to_sequence(qpm=qpm)
assert (generated_sequence.total_time - generate_section.end_time) <= 1e-5
return generated_sequence
def music_generator(primer='erik_gnossienne',
primer_begin_buffer=10,
primer_length=90,
output_path='.',
filename='./public/output'):
SF2_PATH = './models/Yamaha-C5-Salamander-JNv5.1.sf2'
SAMPLE_RATE = 16000
# Upload a MIDI file and convert to NoteSequence.
def upload_midi():
data = list(files.upload().values())
if len(data) > 1:
print('Multiple files uploaded; using only one.')
return mm.midi_to_note_sequence(data[0])
# Decode a list of IDs.
def decode(ids, encoder):
ids = list(ids)
if text_encoder.EOS_ID in ids:
ids = ids[:ids.index(text_encoder.EOS_ID)]
return encoder.decode(ids)
model_name = 'transformer'
hparams_set = 'transformer_tpu'
ckpt_path = './models/checkpoints/unconditional_model_16.ckpt'
class PianoPerformanceLanguageModelProblem(score2perf.Score2PerfProblem):
@property
def add_eos_symbol(self):
return True
problem = PianoPerformanceLanguageModelProblem()
unconditional_encoders = problem.get_feature_encoders()
# Set up HParams.
hparams = trainer_lib.create_hparams(hparams_set=hparams_set)
trainer_lib.add_problem_hparams(hparams, problem)
hparams.num_hidden_layers = 16
hparams.sampling_method = 'random'
# Set up decoding HParams.
decode_hparams = decoding.decode_hparams()
decode_hparams.alpha = 0.0
decode_hparams.beam_size = 1
# Create Estimator.
run_config = trainer_lib.create_run_config(hparams)
estimator = trainer_lib.create_estimator(model_name,
hparams,
run_config,
decode_hparams=decode_hparams)
# These values will be changed by subsequent cells.
targets = []
decode_length = 0
# Create input generator (so we can adjust priming and
# decode length on the fly).
def input_generator():
global targets
global decode_length
while True:
yield {
'targets': np.array([targets], dtype=np.int32),
'decode_length': np.array(decode_length, dtype=np.int32)
}
# Start the Estimator, loading from the specified checkpoint.
input_fn = decoding.make_input_fn_from_generator(input_generator())
unconditional_samples = estimator.predict(input_fn,
checkpoint_path=ckpt_path)
# "Burn" one.
_ = next(unconditional_samples)
filenames = {
'C major arpeggio': './models/primers/c_major_arpeggio.mid',
'C major scale': './models/primers/c_major_scale.mid',
'Clair de Lune': './models/primers/clair_de_lune.mid',
'Classical':
'audio_midi/Classical_Piano_piano-midi.de_MIDIRip/bach/bach_846_format0.mid',
'erik_gymnopedie': 'audio_midi/erik_satie/gymnopedie_1_(c)oguri.mid',
'erik_gymnopedie_2': 'audio_midi/erik_satie/gymnopedie_2_(c)oguri.mid',
'erik_gymnopedie_3': 'audio_midi/erik_satie/gymnopedie_3_(c)oguri.mid',
'erik_gnossienne': 'audio_midi/erik_satie/gnossienne_1_(c)oguri.mid',
'erik_gnossienne_2': 'audio_midi/erik_satie/gnossienne_2_(c)oguri.mid',
'erik_gnossienne_3': 'audio_midi/erik_satie/gnossienne_3_(c)oguri.mid',
'erik_gnossienne_dery':
'audio_midi/erik_satie/gnossienne_1_(c)dery.mid',
'erik_gnossienne_dery_2':
'audio_midi/erik_satie/gnossienne_2_(c)dery.mid',
'erik_gnossienne_dery_3':
'audio_midi/erik_satie/gnossienne_3_(c)dery.mid',
'erik_gnossienne_dery_5':
'audio_midi/erik_satie/gnossienne_5_(c)dery.mid',
'erik_gnossienne_dery_6':
'audio_midi/erik_satie/gnossienne_6_(c)dery.mid',
'1': 'audio_midi/erik_satie/1.mid',
'2': 'audio_midi/erik_satie/2.mid',
'3': 'audio_midi/erik_satie/3.mid',
'4': 'audio_midi/erik_satie/4.mid',
'5': 'audio_midi/erik_satie/5.mid',
'6': 'audio_midi/erik_satie/6.mid',
'7': 'audio_midi/erik_satie/7.mid',
'8': 'audio_midi/erik_satie/8.mid',
'9': 'audio_midi/erik_satie/9.mid',
'10': 'audio_midi/erik_satie/10.mid',
}
# primer = 'C major scale'
#if primer == 'Upload your own!':
# primer_ns = upload_midi()
#else:
# # Use one of the provided primers.
# primer_ns = mm.midi_file_to_note_sequence(filenames[primer])
primer_ns = mm.midi_file_to_note_sequence(filenames[primer])
# Handle sustain pedal in the primer.
primer_ns = mm.apply_sustain_control_changes(primer_ns)
# Trim to desired number of seconds.
max_primer_seconds = primer_length
if primer_ns.total_time > max_primer_seconds:
print('Primer is longer than %d seconds, truncating.' %
max_primer_seconds)
primer_ns = mm.extract_subsequence(
primer_ns, primer_begin_buffer,
max_primer_seconds + primer_begin_buffer)
# Remove drums from primer if present.
if any(note.is_drum for note in primer_ns.notes):
print('Primer contains drums; they will be removed.')
notes = [note for note in primer_ns.notes if not note.is_drum]
del primer_ns.notes[:]
primer_ns.notes.extend(notes)
# Set primer instrument and program.
for note in primer_ns.notes:
note.instrument = 1
note.program = 0
## Play and plot the primer.
#mm.play_sequence(
# primer_ns,
# synth=mm.fluidsynth, sample_rate=SAMPLE_RATE, sf2_path=SF2_PATH)
#mm.plot_sequence(primer_ns)
mm.sequence_proto_to_midi_file(
primer_ns, join(output_path, 'primer_{}.mid'.format(filename)))
targets = unconditional_encoders['targets'].encode_note_sequence(primer_ns)
# Remove the end token from the encoded primer.
targets = targets[:-1]
decode_length = max(0, 10000 - len(targets))
if len(targets) >= 4096:
print(
'Primer has more events than maximum sequence length; nothing will be generated.'
)
# Generate sample events.
sample_ids = next(unconditional_samples)['outputs']
# Decode to NoteSequence.
midi_filename = decode(sample_ids,
encoder=unconditional_encoders['targets'])
ns = mm.midi_file_to_note_sequence(midi_filename)
print('Sample IDs: {}'.format(sample_ids))
print('Sample IDs length: {}'.format(len(sample_ids)))
print('Encoder: {}'.format(unconditional_encoders['targets']))
print('Unconditional Samples: {}'.format(unconditional_samples))
# print('{}'.format(ns))
# continuation_ns = mm.concatenate_sequences([primer_ns, ns])
continuation_ns = ns
# mm.play_sequence(
# continuation_ns,
# synth=mm.fluidsynth, sample_rate=SAMPLE_RATE, sf2_path=SF2_PATH)
# mm.plot_sequence(continuation_ns)
# try:
audio = mm.fluidsynth(continuation_ns,
sample_rate=SAMPLE_RATE,
sf2_path=SF2_PATH)
normalizer = float(np.iinfo(np.int16).max)
array_of_ints = np.array(np.asarray(audio) * normalizer, dtype=np.int16)
wavfile.write(join(output_path, filename + '.wav'), SAMPLE_RATE,
array_of_ints)
print('[+] Output stored as {}'.format(filename + '.wav'))
mm.sequence_proto_to_midi_file(
continuation_ns,
join(output_path, 'continuation_{}.mid'.format(filename)))
def trim_sequence(self, seq, num_seconds=12.0):
seq = mm.extract_subsequence(seq, 0.0, num_seconds)
seq.total_time = num_seconds
