def main(exp, frame_sizes, generate_from, **params):
params = dict(default_params,
exp=exp,
frame_sizes=frame_sizes,
generate_from=generate_from,
**params)
model = SampleRNN(
frame_sizes=params['frame_sizes'],
n_rnn=params['n_rnn'],
dim=params['dim'],
learn_h0=params['learn_h0'],
q_levels=params['q_levels'],
nb_classes=params['nb_classes'],
weight_norm=params['weight_norm'],
)
# model = SampleRNN([16, 4], 2, 1024, True, 256, True)
print('Loading saved model' + params['generate_from'])
checkpoint = torch.load(params['generate_from'])
temporary_dict = {}
for k, v in checkpoint.items():
temporary_dict[k[6:]] = v
checkpoint = temporary_dict
model.load_state_dict(checkpoint)
if not os.path.exists(params['generate_to']):
os.mkdir(params['generate_to'])
print(params['cond'])
generator = GeneratorPlugin(params['generate_to'], params['n_samples'],
params['sample_length'], params['sample_rate'],
params['nb_classes'], params['cond'])
generator.register_generate(model.cuda(), params['cuda'])
generator.epoch(exp)
def main(exp, frame_sizes, dataset, **params):
params = dict(
default_params,
exp=exp, frame_sizes=frame_sizes, dataset=dataset,
**params
)
results_path = setup_results_dir(params)
tee_stdout(os.path.join(results_path, 'log'))
model = SampleRNN(
frame_sizes=params['frame_sizes'],
n_rnn=params['n_rnn'],
dim=params['dim'],
learn_h0=params['learn_h0'],
q_levels=params['q_levels'],
weight_norm=params['weight_norm']
)
predictor = Predictor(model)
if params['cuda']:
model = model.cuda()
predictor = predictor.cuda()
optimizer = gradient_clipping(torch.optim.Adam(predictor.parameters()))
test_split = 1 - params['test_frac']
val_split = test_split - params['val_frac']
data_loader = make_data_loader(model.lookback, params)
train_data_loader = data_loader(0, val_split, eval=False)
val_data_loader = data_loader(val_split, test_split, eval=True)
test_data_loader = data_loader(test_split, 1, eval=True)
def main(exp, frame_sizes, dataset, **params):
params = dict(default_params,
exp=exp,
frame_sizes=frame_sizes,
dataset=dataset,
**params)
results_path = setup_results_dir(params)
tee_stdout(os.path.join(results_path, 'log'))
model = SampleRNN(frame_sizes=params['frame_sizes'],
n_rnn=params['n_rnn'],
dim=params['dim'],
learn_h0=params['learn_h0'],
q_levels=params['q_levels'],
weight_norm=params['weight_norm'])
predictor = Predictor(model)
# if params['cuda']:
# model = model.cuda()
# predictor = predictor.cuda()
#
# # Adam lr = 0.001, betas=(0.9, 0.999) weight_decay=0
test_split = 1 - params['test_frac']
val_split = test_split - params['val_frac']
checkpoints_path = os.path.join(results_path, 'checkpoints')
checkpoint_data = load_last_checkpoint(checkpoints_path, load_best=False)
data_loader = make_data_loader(model.lookback, params)
if checkpoint_data is not None:
(state_dict, epoch, iteration) = checkpoint_data
predictor.load_state_dict(state_dict)
generator = Generator(predictor.model, cuda=True)
else:
raise FileNotFoundError('There is no valid checkpoint.')
save_path = '/mnt/IDMT-WORKSPACE/DATA-STORE/xiaoyg/samplernn/generated_audio'
if not os.path.exists(save_path):
os.makedirs(save_path)
for i in range(0, 10):
y = generator(
1000,
160000,
i,
data_seed=data_loader(val_split, test_split,
eval=True).dataset).float().numpy()
for j in range(len(y)):
path = os.path.join(save_path,
'{}_{}.wav'.format(class_label[i], j))
wav = y[j, :].astype(np.int16)
wav.resize(wav.size, 1)
write_wav(path, wav, 16000, 'PCM_16')
def main(exp, frame_sizes, dataset, **params):
params = dict(default_params,
exp=exp,
frame_sizes=frame_sizes,
dataset=dataset,
**params)
results_path = setup_results_dir(params)
tee_stdout(os.path.join(results_path, 'log'))
model = SampleRNN(frame_sizes=params['frame_sizes'],
n_rnn=params['n_rnn'],
dim=params['dim'],
learn_h0=params['learn_h0'],
q_levels=params['q_levels'],
weight_norm=params['weight_norm'])
predictor = Predictor(model)
if params['cuda']:
model = model.cuda()
predictor = predictor.cuda()
optimizer = gradient_clipping(torch.optim.Adam(predictor.parameters()))
data_loader = make_data_loader(model.lookback, params)
test_split = 1 - params['test_frac']
val_split = test_split - params['val_frac']
checkpoints_path = os.path.join(results_path, 'checkpoints')
samples_path = os.path.join(results_path, 'samples')
trainer = Trainer(predictor,
sequence_nll_loss_bits,
optimizer,
data_loader(0, val_split, eval=False),
checkpoints_path,
samples_path,
params['n_samples'],
params['sample_length'],
params['sample_rate'],
data_loader(val_split, test_split, eval=True),
data_loader(test_split, 1, eval=True),
cuda=params['cuda'])
checkpoints_path = os.path.join(results_path, 'checkpoints')
checkpoint_data = load_last_checkpoint(checkpoints_path)
if checkpoint_data is not None:
(state_dict, epoch, iteration) = checkpoint_data
trainer.epochs = epoch
trainer.iterations = iteration
predictor.load_state_dict(state_dict)
trainer.run(params['epoch_limit'])
def main():
model = SampleRNN(
frame_sizes=[16, 4], n_rnn=1, dim=1024, learn_h0=True, q_levels=256
)
predictor = Predictor(model).cuda()
predictor.load_state_dict(torch.load('model.tar'))
generator = Generator(predictor.model, cuda=True)
t = time()
samples = generator(5, 16000)
print('generated in {}s'.format(time() - t))
write_wav(
'sample.wav',
samples.cpu().float().numpy()[0, :],
sr=16000,
norm=True
)
def main(exp, frame_sizes, dataset, **params):
params = dict(default_params,
exp=exp,
frame_sizes=frame_sizes,
dataset=dataset,
**params)
results_path = setup_results_dir(params)
tee_stdout(os.path.join(results_path, 'log'))
# Save samplernn parameters in .json for future audio generation
import json
with open(os.path.join(results_path, 'sample_rnn_params.json'), 'w') as fp:
json.dump(params, fp, sort_keys=True, indent=4)
model = SampleRNN(frame_sizes=params['frame_sizes'],
n_rnn=params['n_rnn'],
dim=params['dim'],
learn_h0=params['learn_h0'],
q_levels=params['q_levels'],
weight_norm=params['weight_norm'])
predictor = Predictor(model)
if params['cuda']:
model = model.cuda()
predictor = predictor.cuda()
optimizer = gradient_clipping(torch.optim.Adam(predictor.parameters()))
data_loader = make_data_loader(model.lookback, params)
test_split = 1 - params['test_frac']
val_split = test_split - params['val_frac']
trainer = Trainer(predictor,
sequence_nll_loss_bits,
optimizer,
data_loader(0, val_split, eval=False),
cuda=params['cuda'])
checkpoints_path = os.path.join(results_path, 'checkpoints')
checkpoint_data = load_last_checkpoint(checkpoints_path)
if checkpoint_data is not None:
(state_dict, epoch, iteration) = checkpoint_data
trainer.epochs = epoch
trainer.iterations = iteration
predictor.load_state_dict(state_dict)
else:
trainer.epochs = 0
trainer.iterations = 0
torch.save(predictor,
os.path.join(checkpoints_path, "pytorch_model.bin"))
# else:
# print("***** Saving fine-tuned model *****")
# output_model_file = os.path.join(params['results_path'], "pytorch_model.bin")
# if params['cuda']:
# torch.save(predictor, output_model_file)
# else:
# torch.save(predictor, output_model_file)
trainer.register_plugin(
TrainingLossMonitor(smoothing=params['loss_smoothing']))
trainer.register_plugin(
ValidationPlugin(data_loader(val_split, test_split, eval=True),
data_loader(test_split, 1, eval=True)))
trainer.register_plugin(AbsoluteTimeMonitor())
trainer.register_plugin(
SaverPlugin(checkpoints_path, params['keep_old_checkpoints']))
trainer.register_plugin(
GeneratorPlugin(os.path.join(results_path, 'samples'),
params['n_samples'], params['sample_length'],
params['sample_rate'], params['sampling_temperature']))
trainer.register_plugin(
Logger(['training_loss', 'validation_loss', 'test_loss', 'time']))
trainer.register_plugin(
StatsPlugin(results_path,
iteration_fields=[
'training_loss', ('training_loss', 'running_avg'),
'time'
],
epoch_fields=['validation_loss', 'test_loss', 'time'],
plots={
'loss': {
'x':
'iteration',
'ys': [
'training_loss',
('training_loss', 'running_avg'),
'validation_loss',
'test_loss',
],
'log_y':
True
}
}))
init_comet(params, trainer)
trainer.run(params['epoch_limit'])
def main(exp, frame_sizes, dataset, **params):
params = dict(default_params,
exp=exp,
frame_sizes=frame_sizes,
dataset=dataset,
**params)
os.environ['CUDA_VISIBLE_DEVICES'] = params['gpu']
results_path = setup_results_dir(params)
tee_stdout(os.path.join(results_path, 'log'))
model = SampleRNN(frame_sizes=params['frame_sizes'],
n_rnn=params['n_rnn'],
dim=params['dim'],
learn_h0=params['learn_h0'],
q_levels=params['q_levels'],
weight_norm=params['weight_norm'],
dropout=params['dropout'])
predictor = Predictor(model)
if params['cuda']:
model = model.cuda()
predictor = predictor.cuda()
optimizer = gradient_clipping(
torch.optim.Adam(predictor.parameters(), lr=params['lr']))
data_loader = make_data_loader(model.lookback, params)
test_split = 1 - params['test_frac']
val_split = test_split - params['val_frac']
criterion = sequence_nll_loss_bits
checkpoints_path = os.path.join(results_path, 'checkpoints')
checkpoint_data = load_last_checkpoint(checkpoints_path, params)
if checkpoint_data is not None:
(state_dict, epoch, iteration) = checkpoint_data
start_epoch = int(epoch)
global_step = iteration
start_epoch = iteration
predictor.load_state_dict(state_dict)
else:
start_epoch = 0
global_step = 0
#writer = SummaryWriter("runs/{}-{}".format(params['dataset'], str(datetime.datetime.now()).split('.')[0].replace(' ', '-')))
writer = SummaryWriter(
os.path.join(
results_path, "{}-{}".format(
params['dataset'],
str(datetime.datetime.now()).split('.')[0].replace(' ', '-'))))
dataset_train = data_loader(0, val_split, eval=False)
dataset_val = data_loader(val_split, test_split, eval=True)
dataset_test = data_loader(test_split, 1, eval=True)
generator = Generator(predictor.model, params['cuda'])
best_val_loss = 10000000000000
for e in range(start_epoch, int(params['epoch_limit'])):
for i, data in enumerate(dataset_train):
batch_inputs = data[:-1]
batch_target = data[-1]
def wrap(input):
if torch.is_tensor(input):
input = torch.autograd.Variable(input)
if params['cuda']:
input = input.cuda()
return input
batch_inputs = list(map(wrap, batch_inputs))
batch_target = torch.autograd.Variable(batch_target)
if params['cuda']:
batch_target = batch_target.cuda()
plugin_data = [None, None]
def closure():
batch_output = predictor(*batch_inputs)
loss = criterion(batch_output, batch_target)
loss.backward()
if plugin_data[0] is None:
plugin_data[0] = batch_output.data
plugin_data[1] = loss.data
return loss
optimizer.zero_grad()
optimizer.step(closure)
train_loss = plugin_data[1]
# stats: iteration
writer.add_scalar('train/train loss', train_loss, global_step)
print("E:{:03d}-S{:05d}: Loss={}".format(e, i, train_loss))
global_step += 1
# validation: per epoch
predictor.eval()
with torch.no_grad():
loss_sum = 0
n_examples = 0
for data in dataset_val:
batch_inputs = data[:-1]
batch_target = data[-1]
batch_size = batch_target.size()[0]
def wrap(input):
if torch.is_tensor(input):
input = torch.autograd.Variable(input)
if params['cuda']:
input = input.cuda()
return input
batch_inputs = list(map(wrap, batch_inputs))
batch_target = torch.autograd.Variable(batch_target)
if params['cuda']:
batch_target = batch_target.cuda()
batch_output = predictor(*batch_inputs)
loss_sum += criterion(batch_output,
batch_target).item() * batch_size
n_examples += batch_size
val_loss = loss_sum / n_examples
writer.add_scalar('validation/validation loss', val_loss,
global_step)
print("== Validation Step E:{:03d}: Loss={} ==".format(
e, val_loss))
predictor.train()
# saver: epoch
last_pattern = 'ep{}-it{}'
best_pattern = 'best-ep{}-it{}'
if not params['keep_old_checkpoints']:
pattern = os.path.join(checkpoints_path,
last_pattern.format('*', '*'))
for file_name in glob(pattern):
os.remove(file_name)
torch.save(
predictor.state_dict(),
os.path.join(checkpoints_path, last_pattern.format(e,
global_step)))
cur_val_loss = val_loss
if cur_val_loss < best_val_loss:
pattern = os.path.join(checkpoints_path,
last_pattern.format('*', '*'))
for file_name in glob(pattern):
os.remove(file_name)
torch.save(
predictor.state_dict(),
os.path.join(checkpoints_path,
best_pattern.format(e, global_step)))
best_val_loss = cur_val_loss
generate_sample(generator, params, writer, global_step, results_path,
e)
# generate final results
generate_sample(generator, params, None, global_step, results_path, 0)
def main(exp, frame_sizes, dataset, **params):
params = dict(default_params,
exp=exp,
frame_sizes=frame_sizes,
dataset=dataset,
**params)
results_path = setup_results_dir(params)
tee_stdout(os.path.join(results_path, 'log'))
model = SampleRNN(frame_sizes=params['frame_sizes'],
n_rnn=params['n_rnn'],
dim=params['dim'],
learn_h0=params['learn_h0'],
q_levels=params['q_levels'],
weight_norm=params['weight_norm'])
predictor = Predictor(model)
if params['cuda']:
model = model.cuda()
predictor = predictor.cuda()
optimizer = gradient_clipping(torch.optim.Adam(predictor.parameters()))
data_loader = make_data_loader(model.lookback, params)
test_split = 1 - params['test_frac']
val_split = test_split - params['val_frac']
trainer = Trainer(predictor,
sequence_nll_loss_bits,
optimizer,
data_loader(0, val_split, eval=False),
cuda=params['cuda'])
checkpoints_path = os.path.join(results_path, 'checkpoints')
checkpoint_data = load_last_checkpoint(checkpoints_path)
if checkpoint_data is not None:
(state_dict, epoch, iteration) = checkpoint_data
trainer.epochs = epoch
trainer.iterations = iteration
predictor.load_state_dict(state_dict)
trainer.register_plugin(
TrainingLossMonitor(smoothing=params['loss_smoothing']))
trainer.register_plugin(
ValidationPlugin(data_loader(val_split, test_split, eval=True),
data_loader(test_split, 1, eval=True)))
trainer.register_plugin(AbsoluteTimeMonitor())
trainer.register_plugin(
SaverPlugin(checkpoints_path, params['keep_old_checkpoints']))
trainer.register_plugin(
GeneratorPlugin(os.path.join(results_path,
'samples'), params['n_samples'],
params['sample_length'], params['sample_rate']))
trainer.register_plugin(
Logger(['training_loss', 'validation_loss', 'test_loss', 'time']))
trainer.register_plugin(
StatsPlugin(results_path,
iteration_fields=[
'training_loss', ('training_loss', 'running_avg'),
'time'
],
epoch_fields=['validation_loss', 'test_loss', 'time'],
plots={
'loss': {
'x':
'iteration',
'ys': [
'training_loss',
('training_loss', 'running_avg'),
'validation_loss',
'test_loss',
],
'log_y':
True
}
}))
init_comet(params, trainer)
trainer.run(params['epoch_limit'])
def main(exp, dataset, **params):
params = dict(default_params, exp=exp, dataset=dataset, **params)
print(params)
storage_client = None
bucket = None
path = os.path.join(params['datasets_path'], params['dataset'])
if params['bucket']:
storage_client = storage.Client()
bucket = Bucket(storage_client, params['bucket'])
preload_dataset(path, storage_client, bucket)
results_path = setup_results_dir(params)
tee_stdout(os.path.join(results_path, 'log'))
(quantize, dequantize) = quantizer(params['q_method'])
model = SampleRNN(frame_sizes=params['frame_sizes'],
n_rnn=params['n_rnn'],
dim=params['dim'],
learn_h0=params['learn_h0'],
q_levels=params['q_levels'],
weight_norm=params['weight_norm'])
predictor = Predictor(model, dequantize)
if params['cuda'] is not False:
print(params['cuda'])
model = model.cuda()
predictor = predictor.cuda()
optimizer = gradient_clipping(
torch.optim.Adam(predictor.parameters(), lr=params['learning_rate']))
data_loader = make_data_loader(path, model.lookback, quantize, params)
test_split = 1 - params['test_frac']
val_split = test_split - params['val_frac']
trainer = Trainer(predictor,
sequence_nll_loss_bits,
optimizer,
data_loader(0, val_split, eval=False),
cuda=params['cuda'])
checkpoints_path = os.path.join(results_path, 'checkpoints')
checkpoint_data = load_last_checkpoint(checkpoints_path, storage_client,
bucket)
if checkpoint_data is not None:
(state_dict, epoch, iteration) = checkpoint_data
trainer.epochs = epoch
trainer.iterations = iteration
predictor.load_state_dict(state_dict)
trainer.register_plugin(
TrainingLossMonitor(smoothing=params['loss_smoothing']))
trainer.register_plugin(
ValidationPlugin(data_loader(val_split, test_split, eval=True),
data_loader(test_split, 1, eval=True)))
trainer.register_plugin(SchedulerPlugin(params['lr_scheduler_step']))
def upload(file_path):
if bucket is None:
return
name = file_path.replace(os.path.abspath(os.curdir) + '/', '')
blob = Blob(name, bucket)
try:
blob.upload_from_filename(file_path, timeout=300)
except Exception as e:
print(str(e))
trainer.register_plugin(AbsoluteTimeMonitor())
samples_path = os.path.join(results_path, 'samples')
trainer.register_plugin(
SaverPlugin(checkpoints_path, params['keep_old_checkpoints'], upload))
trainer.register_plugin(
GeneratorPlugin(samples_path,
params['n_samples'],
params['sample_length'],
params['sample_rate'],
params['q_levels'],
dequantize,
params['sampling_temperature'],
upload=upload))
trainer.register_plugin(
Logger(['training_loss', 'validation_loss', 'test_loss', 'time']))
trainer.register_plugin(
StatsPlugin(
results_path,
iteration_fields=[
'training_loss',
#('training_loss', 'running_avg'),
'time'
],
epoch_fields=[
'training_loss', ('training_loss', 'running_avg'),
'validation_loss', 'test_loss', 'time'
],
plots={
'loss': {
'x':
'iteration',
'ys': [
'training_loss',
# ('training_loss', 'running_avg'),
'validation_loss',
'test_loss'
],
'log_y':
True
}
}))
init_comet(params, trainer, samples_path, params['n_samples'],
params['sample_rate'])
trainer.run(params['epoch_limit'])
def main(checkpoint, **args):
task_id = setup_logging(
'gen', logging.NOTSET if args.get('debug', False) else logging.INFO)
params = dict(
{
'n_rnn': 3,
'dim': 1024,
'learn_h0': False,
'q_levels': 256,
'weight_norm': True,
'frame_sizes': [16, 16, 4],
'sample_rate': 16000,
'n_samples': 1,
'sample_length': 16000 * 60 * 4,
'sampling_temperature': 1,
'q_method': QMethod.LINEAR,
},
exp=checkpoint,
**args)
logging.info(str(params))
logging.info('booting')
# dataset = storage_client.list_blobs(bucket, prefix=path)
# for blob in dataset:
# blob.download_to_filename(blob.name)
bucket = None
if args['bucket']:
logging.debug('setup google storage bucket {}'.format(args['bucket']))
storage_client = storage.Client()
bucket = Bucket(storage_client, args['bucket'])
preload_checkpoint(checkpoint, storage_client, bucket)
results_path = os.path.abspath(
os.path.join(checkpoint, os.pardir, os.pardir, task_id))
ensure_dir_exists(results_path)
checkpoint = os.path.abspath(checkpoint)
tmp_pretrained_state = torch.load(
checkpoint,
map_location=lambda storage, loc: storage.cuda(0)
if args['cuda'] else storage)
# Load all tensors onto GPU 1
# torch.load('tensors.pt', map_location=lambda storage, loc: storage.cuda(1))
pretrained_state = OrderedDict()
for k, v in tmp_pretrained_state.items():
# Delete "model." from key names since loading the checkpoint automatically attaches it
layer_name = k.replace("model.", "")
pretrained_state[layer_name] = v
# print("k: {}, layer_name: {}, v: {}".format(k, layer_name, np.shape(v)))
# Create model with same parameters as used in training
model = SampleRNN(frame_sizes=params['frame_sizes'],
n_rnn=params['n_rnn'],
dim=params['dim'],
learn_h0=params['learn_h0'],
q_levels=params['q_levels'],
weight_norm=params['weight_norm'])
if params['cuda']:
model = model.cuda()
# Load pretrained model
model.load_state_dict(pretrained_state)
def upload(file_path):
if bucket is None:
return
# remove prefix /app
name = file_path.replace(os.path.abspath(os.curdir) + '/', '')
blob = Blob(name, bucket)
logging.info('uploading {}'.format(name))
blob.upload_from_filename(file_path)
(_, dequantize) = quantizer(params['q_method'])
gen = Gen(Runner(model), params['cuda'])
gen.register_plugin(
GeneratorPlugin(results_path, params['n_samples'],
params['sample_length'], params['sample_rate'],
params['q_levels'], dequantize,
params['sampling_temperature'], upload))
gen.run()
# RESULTS_PATH = 'results/exp:TEST-frame_sizes:16,4-n_rnn:2-dataset:piano3/'
# PRETRAINED_PATH = RESULTS_PATH + 'checkpoints/best-ep21-it29610'
GENERATED_PATH = RESULTS_PATH + 'generated/'
if not os.path.exists(GENERATED_PATH):
os.mkdir(GENERATED_PATH)
# Load model parameters from .json for audio generation
params_path = RESULTS_PATH + 'sample_rnn_params.json'
with open(params_path, 'r') as fp:
params = json.load(fp)
# Create model with same parameters as used in training
model = SampleRNN(
frame_sizes=params['frame_sizes'],
n_rnn=params['n_rnn'],
dim=params['dim'],
learn_h0=params['learn_h0'],
q_levels=params['q_levels'],
weight_norm=params['weight_norm']
)
#model = model.cuda()
# Delete "model." from key names since loading the checkpoint automatically attaches it to the key names
pretrained_state = torch.load(PRETRAINED_PATH)
new_pretrained_state = OrderedDict()
for k, v in pretrained_state.items():
layer_name = k.replace("model.", "")
new_pretrained_state[layer_name] = v
# print("k: {}, layer_name: {}, v: {}".format(k, layer_name, np.shape(v)))
# Load pretrained model
def main(frame_sizes, **params):
use_cuda = torch.cuda.is_available()
params = dict(default_params, frame_sizes=frame_sizes, **params)
# Redefine parameters listed in the experiment directory and separated with '~'
for i in params['model'].split('/')[1].split('~'):
param = i.split(':')
if param[0] in params:
params[param[0]] = as_type(param[1], type(params[param[0]]))
# Define npy file names with maximum and minimum values of de-normalized conditioners
npy_name_min_max_cond = 'npy_datasets/min_max' + params['norm_ind'] * '_ind' + (not params['norm_ind']) * '_joint' \
+ params['static_spk'] * '_static' + '.npy'
# Define npy file name with array of unique speakers in dataset
npy_name_spk_id = 'npy_datasets/spk_id.npy'
# Get file names from partition's list
file_names = open(
str(params['datasets_path']) + 'generate_cond_gina.list',
'r').read().splitlines()
spk_names = open(
str(params['datasets_path']) + 'generate_spk_gina.list',
'r').read().splitlines()
datasets_path = os.path.join(params['datasets_path'], params['cond_set'])
spk = np.load(npy_name_spk_id)
if len(spk_names) != len(file_names):
print(
'Length of speaker file do not match length of conditioner file.')
quit()
print('Generating', len(file_names), 'audio files')
for i in range(len(file_names)):
print('Generating Audio', i)
print('Generating...', file_names[i])
# Load CC conditioner
c = np.loadtxt(datasets_path + file_names[i] + '.cc')
# Load LF0 conditioner
f0file = np.loadtxt(datasets_path + file_names[i] + '.lf0')
f0, _ = interpolation(f0file, -10000000000)
f0 = f0.reshape(f0.shape[0], 1)
# Load FV conditioner
fvfile = np.loadtxt(datasets_path + file_names[i] + '.gv')
fv, uv = interpolation(fvfile, 1e3)
num_fv = fv.shape[0]
uv = uv.reshape(num_fv, 1)
fv = fv.reshape(num_fv, 1)
# Load speaker conditioner
speaker = np.where(spk == spk_names[i])[0][0]
cond = np.concatenate((c, f0), axis=1)
cond = np.concatenate((cond, fv), axis=1)
cond = np.concatenate((cond, uv), axis=1)
# Load maximum and minimum of de-normalized conditioners
min_cond = np.load(npy_name_min_max_cond)[0]
max_cond = np.load(npy_name_min_max_cond)[1]
# Normalize conditioners with absolute maximum and minimum for each speaker of training partition
if params['norm_ind']:
print(
'Normalizing conditioners for each speaker of training dataset'
)
cond = (cond - min_cond[speaker]) / (max_cond[speaker] -
min_cond[speaker])
else:
print('Normalizing conditioners jointly')
cond = (cond - min_cond) / (max_cond - min_cond)
print('Shape cond', cond.shape)
if params['look_ahead']:
delayed = np.copy(cond)
delayed[:-1, :] = delayed[1:, :]
cond = np.concatenate((cond, delayed), axis=1)
print('Shape cond after look ahead', cond.shape)
print(cond.shape)
seed = params.get('seed')
init_random_seed(seed, use_cuda)
spk_dim = len([
i for i in os.listdir(
os.path.join(params['datasets_path'], params['cond_set']))
if os.path.islink(
os.path.join(params['datasets_path'], params['cond_set']) +
'/' + i)
])
print('Start Generate SampleRNN')
model = SampleRNN(frame_sizes=params['frame_sizes'],
n_rnn=params['n_rnn'],
dim=params['dim'],
learn_h0=params['learn_h0'],
q_levels=params['q_levels'],
ulaw=params['ulaw'],
weight_norm=params['weight_norm'],
cond_dim=params['cond_dim'] *
(1 + params['look_ahead']),
spk_dim=spk_dim,
qrnn=params['qrnn'])
print(model)
if use_cuda:
model = model.cuda()
predictor = Predictor(model).cuda()
else:
predictor = Predictor(model)
f_name = params['model']
model_data = load_model(f_name)
if model_data is None:
sys.exit('ERROR: Model not found in' + str(f_name))
(state_dict, epoch_index, iteration) = model_data
print('OK: Read model', f_name, '(epoch:', epoch_index, ')')
print(state_dict)
predictor.load_state_dict(state_dict)
original_name = file_names[i].split('/')[1]
if original_name == "..":
original_name = file_names[i].split('/')[3]
generator = RunGenerator(model=model,
sample_rate=params['sample_rate'],
cuda=use_cuda,
epoch=epoch_index,
cond=cond,
spk_list=spk,
speaker=speaker,
checkpoints_path=f_name,
original_name=original_name)
generator(params['n_samples'], params['sample_length'], cond, speaker)
def main(exp, frame_sizes, dataset, **params):
print('main called')
params = dict(
default_params,
# exp='TEST', frame_sizes=[16,4], dataset='piano',
exp=exp,
frame_sizes=frame_sizes,
dataset=dataset,
**params)
results_path = setup_results_dir(params)
tee_stdout(os.path.join(results_path, 'log'))
model = SampleRNN(frame_sizes=params['frame_sizes'],
n_rnn=params['n_rnn'],
dim=params['dim'],
learn_h0=params['learn_h0'],
q_levels=params['q_levels'],
weight_norm=params['weight_norm'])
predictor = Predictor(model)
if params['cuda']:
model = model.cuda()
predictor = predictor.cuda()
optimizer = torch.optim.Adam(predictor.parameters())
# optimizer = gradient_clipping(torch.optim.Adam(predictor.parameters()))
loss = sequence_nll_loss_bits
test_split = 1 - params['test_frac']
val_split = test_split - params['val_frac']
data_loader = make_data_loader(model.lookback, params)
train_loader = data_loader(0, val_split, eval=False)
val_loader = data_loader(val_split, test_split, eval=True)
test_loader = data_loader(test_split, 1, eval=True)
trainer = create_supervised_trainer(predictor, optimizer, loss,
params['cuda'])
evaluator = create_supervised_evaluator(predictor,
metrics={'nll': Loss(loss)})
# @trainer.on(Events.ITERATION_COMPLETED)
# def log_training_loss(trainer):
# print("Epoch[{}] Loss: {:.2f}".format(
# trainer.state.epoch, trainer.state.output))
@trainer.on(Events.EPOCH_COMPLETED)
def log_training_results(trainer):
evaluator.run(train_loader)
metrics = evaluator.state.metrics
print("Training Results - Epoch: {} Avg loss: {:.2f}".format(
trainer.state.epoch, metrics['nll']))
@trainer.on(Events.EPOCH_COMPLETED)
def log_validation_results(trainer):
evaluator.run(val_loader)
metrics = evaluator.state.metrics
print("Validation Results - Epoch: {} Avg loss: {:.2f}".format(
trainer.state.epoch, metrics['nll']))
trainer.run(train_loader, max_epochs=2)
print('train complete!')
def main(exp, frame_sizes, dataset, **params):
params = dict(
default_params,
exp=exp, frame_sizes=frame_sizes, dataset=dataset,
**params
)
results_path = setup_results_dir(params)
tee_stdout(os.path.join(results_path, 'log'))
# Save samplernn parameters in .json for future audio generation
import json
with open(os.path.join(results_path, 'sample_rnn_params.json'), 'w') as fp:
json.dump(params, fp, sort_keys=True, indent=4)
# Model
model = SampleRNN(
frame_sizes=params['frame_sizes'],
n_rnn=params['n_rnn'],
dim=params['dim'],
learn_h0=params['learn_h0'],
q_levels=params['q_levels'],
weight_norm=params['weight_norm'],
batch_size=params['batch_size']
)
print("CUDA num: {}".format(torch.cuda.device_count()))
predictor = Predictor(model)
if params['cuda']:
model = model.cuda()
predictor = predictor.cuda()
model_cnnseq2sample = CNNSeq2SampleRNN(params).cuda()
optimizer = gradient_clipping(torch.optim.Adam(predictor.parameters()))
data_loader = make_data_loader(model.lookback, params)
data_loader_test = make_data_loader(model.lookback, params, npz_filename=params['npz_filename_test'])
# test_split = 1 - params['test_frac']
# val_split = test_split - params['val_frac']
trainer = Trainer(
predictor, model_cnnseq2sample, sequence_nll_loss_bits, optimizer,
# data_loader(0, val_split, eval=False),
data_loader(0, 1, eval=False),
cuda=params['cuda']
)
checkpoints_path = os.path.join(results_path, 'checkpoints')
checkpoint_data = load_last_checkpoint(checkpoints_path)
checkpoint_data_cnnseq2sample = load_last_checkpoint(checkpoints_path, model_type='cnnseq2sample')
if checkpoint_data is not None:
(state_dict, epoch, iteration) = checkpoint_data
(state_dict_cnnseq2sample, epoch, iteration) = checkpoint_data_cnnseq2sample
trainer.epochs = epoch
trainer.iterations = iteration
predictor.load_state_dict(state_dict)
model_cnnseq2sample.load_state_dict(state_dict_cnnseq2sample)
trainer.register_plugin(TrainingLossMonitor(
smoothing=params['loss_smoothing']
))
trainer.register_plugin(ValidationPlugin(
# data_loader(val_split, test_split, eval=True),
# data_loader_test(0, 1, eval=True)
data_loader_test(0, params['val_frac'], eval=True),
data_loader_test(params['val_frac'], 1, eval=True)
# data_loader(test_split, 1, eval=True)
))
trainer.register_plugin(AbsoluteTimeMonitor())
trainer.register_plugin(SaverPlugin(
checkpoints_path, params['keep_old_checkpoints']
))
trainer.register_plugin(GeneratorPlugin(
os.path.join(results_path, 'samples'), params['n_samples'],
params['sample_length'], params['sample_rate']
))
trainer.register_plugin(
Logger([
'training_loss',
'validation_loss',
'test_loss',
'time'
])
)
trainer.register_plugin(StatsPlugin(
results_path,
iteration_fields=[
'training_loss',
('training_loss', 'running_avg'),
'time'
],
epoch_fields=[
'validation_loss',
'test_loss',
'time'
],
plots={
'loss': {
'x': 'iteration',
'ys': [
'training_loss',
('training_loss', 'running_avg'),
'validation_loss',
'test_loss',
],
'log_y': True
}
}
))
trainer.run(params['epoch_limit'])
modelname = "ep1-it625"
audio_pref = "r001_{}"
save_raw = False
n_samples = 1
sample_length = 800
sample_rate = 16000
samples_path = os.path.join(modeldir, "gen_samples")
os.makedirs(samples_path, exist_ok=True)
# sys.stderr.write("available models are: {}".format(listdir(modeldir)))
modelpath = os.path.join(modeldir, modelname)
srnn_model1 = SampleRNN(frame_sizes=[4, 16],
n_rnn=2,
dim=1024,
learn_h0=True,
q_levels=256,
weight_norm=True)
if torch.cuda.is_available():
srnn_model1 = srnn_model1.cuda()
predictor1 = Predictor(srnn_model1)
if torch.cuda.is_available():
predictor1 = predictor1.cuda()
if torch.cuda.is_available():
predictor1.load_state_dict(torch.load(modelpath)['model'])
else:
predictor1.load_state_dict(
def main(exp, frame_sizes, dataset, **params):
scheduler = True
use_cuda = torch.cuda.is_available()
print('Start Sample-RNN')
params = dict(
default_params,
exp=exp, frame_sizes=frame_sizes, dataset=dataset,
**params
)
seed = params.get('seed')
init_random_seed(seed, use_cuda)
results_path = setup_results_dir(params)
tee_stdout(os.path.join(results_path, 'log'))
spk_dim = len([i for i in os.listdir(os.path.join(params['datasets_path'], params['dataset']))
if os.path.islink(os.path.join(params['datasets_path'], params['dataset']) + '/' + i)])
print('Create model')
model = SampleRNN(
frame_sizes=params['frame_sizes'],
n_rnn=params['n_rnn'],
dim=params['dim'],
learn_h0=params['learn_h0'],
q_levels=params['q_levels'],
ulaw=params['ulaw'],
weight_norm=params['weight_norm'],
cond_dim=params['cond_dim']*(1+params['look_ahead']),
spk_dim=spk_dim,
qrnn=params['qrnn']
)
if use_cuda:
model = model.cuda()
predictor = Predictor(model).cuda()
else:
predictor = Predictor(model)
print('Done!')
f_name = params['model']
if f_name is not None:
print('pre train with', f_name)
model_data = load_model(f_name)
if model_data is None:
sys.exit('ERROR: Model not found in' + str(f_name))
(state_dict, epoch_index, iteration) = model_data
print('OK: Read model', f_name, '(epoch:', epoch_index, ')')
print(state_dict)
predictor.load_state_dict(state_dict)
print('predictor', predictor)
for name, param in predictor.named_parameters():
print(name, param.size())
optimizer = torch.optim.Adam(predictor.parameters(), lr=params['learning_rate'])
if params['scheduler']:
scheduler = MultiStepLR(optimizer, milestones=[15, 35], gamma=0.1)
optimizer = gradient_clipping(optimizer)
print('Saving results in path', results_path)
print('Read data')
data_loader = make_data_loader(model.lookback, params)
print('Done!')
data_model = data_loader('train')
show_dataset = False
if show_dataset:
for i, full in enumerate(data_model):
print('Data Loader---------------------------------------')
print('batch', i)
(data, reset, target, cond) = full
print('Data', data.size())
print('Target', target.size())
if not params['scheduler']:
scheduler = None
if use_cuda:
cuda = True
else:
cuda = False
writer = SummaryWriter(log_dir='sample_board')
trainer = Trainer(
predictor, sequence_nll_loss_bits, optimizer, data_model, cuda, writer, scheduler
)
checkpoints_path = os.path.join(results_path, 'checkpoints')
checkpoint_data = load_last_checkpoint(checkpoints_path)
if checkpoint_data is not None:
(state_dict, epoch, iteration) = checkpoint_data
trainer.epochs = epoch
trainer.iterations = iteration
predictor.load_state_dict(state_dict)
trainer.register_plugin(TrainingLossMonitor(
smoothing=params['loss_smoothing']
))
trainer.register_plugin(ValidationPlugin(
data_loader('validation'),
data_loader('test'),
writer
))
trainer.register_plugin(AbsoluteTimeMonitor())
trainer.register_plugin(SaverPlugin(
checkpoints_path, params['keep_old_checkpoints']
))
trainer.register_plugin(
Logger([
'training_loss',
'validation_loss',
'test_loss',
'time'
])
)
trainer.register_plugin(StatsPlugin(
results_path,
iteration_fields=[
'training_loss',
('training_loss', 'running_avg'),
'time'
],
epoch_fields=[
'validation_loss',
'test_loss',
'time'
],
plots={
'loss': {
'x': 'iteration',
'ys': [
'training_loss',
('training_loss', 'running_avg'),
'validation_loss',
'test_loss',
],
'log_y': True
}
}
))
trainer.run(params['epoch_limit'])