def phoneme_to_sequence(text,
cleaner_names,
language,
enable_eos_bos=False,
tp=None,
add_blank=False):
# pylint: disable=global-statement
global _phonemes_to_id, _phonemes
if tp:
_, _phonemes = make_symbols(**tp)
_phonemes_to_id = {s: i for i, s in enumerate(_phonemes)}
sequence = []
clean_text = _clean_text(text, cleaner_names)
to_phonemes = text2phone(clean_text, language)
if to_phonemes is None:
print("!! After phoneme conversion the result is None. -- {} ".format(
clean_text))
# iterate by skipping empty strings - NOTE: might be useful to keep it to have a better intonation.
for phoneme in filter(None, to_phonemes.split('|')):
sequence += _phoneme_to_sequence(phoneme)
# Append EOS char
if enable_eos_bos:
sequence = pad_with_eos_bos(sequence, tp=tp)
if add_blank:
sequence = intersperse(
sequence, len(_phonemes)
) # add a blank token (new), whose id number is len(_phonemes)
return sequence
def pad_with_eos_bos(phoneme_sequence, tp=None):
# pylint: disable=global-statement
global _phonemes_to_id, _bos, _eos
if tp:
_bos = tp['bos']
_eos = tp['eos']
_, _phonemes = make_symbols(**tp)
_phonemes_to_id = {s: i for i, s in enumerate(_phonemes)}
return [_phonemes_to_id[_bos]
] + list(phoneme_sequence) + [_phonemes_to_id[_eos]]
def __init__(self,
model_path,
vocoder_path,
encoder_path,
model_config,
vocoder_config,
encoder_config,
speaker_json,
use_cuda=True):
self.speakers_mapping = json.load(open(speaker_json, 'r'))
num_speakers = len(self.speakers_mapping)
self.model_config = load_config(model_config)
self.model_config.forward_attn_mask = True
symbols = ()
phonemes = ()
if 'characters' in self.model_config.keys():
symbols, phonemes = make_symbols(**self.model_config.characters)
self.ap = AudioProcessor(**self.model_config.audio)
num_chars = len(
phonemes) if self.model_config.use_phonemes else len(symbols) - 1
self.model = setup_model(num_chars, num_speakers, self.model_config,
256)
cp = torch.load(model_path, map_location=torch.device('cpu'))
self.model.load_state_dict(cp['model'])
self.model.eval()
self.model.decoder.set_r(cp['r'])
if use_cuda:
self.model.cuda()
self.vocoder_config = load_config(vocoder_config)
if vocoder_path != "":
self.vocoder = setup_generator(self.vocoder_config)
self.vocoder.load_state_dict(
torch.load(vocoder_path, map_location="cpu")["model"])
self.vocoder.remove_weight_norm()
self.vocoder.eval()
self.vocoder.cuda()
else:
self.vocoder = None
self.use_griffin_lim = vocoder_path == ""
self.encoder_config = load_config(encoder_config)
self.encoder = SpeakerEncoder(**self.encoder_config.model)
self.encoder.load_state_dict(torch.load(encoder_path)['model'])
self.encoder.eval()
if use_cuda:
self.encoder.cuda()
def sequence_to_phoneme(sequence, tp=None, add_blank=False):
# pylint: disable=global-statement
'''Converts a sequence of IDs back to a string'''
global _id_to_phonemes, _phonemes
if add_blank:
sequence = list(filter(lambda x: x != len(_phonemes), sequence))
result = ''
if tp:
_, _phonemes = make_symbols(**tp)
_id_to_phonemes = {i: s for i, s in enumerate(_phonemes)}
for symbol_id in sequence:
if symbol_id in _id_to_phonemes:
s = _id_to_phonemes[symbol_id]
result += s
return result.replace('}{', ' ')
def sequence_to_text(sequence, tp=None, add_blank=False):
'''Converts a sequence of IDs back to a string'''
# pylint: disable=global-statement
global _id_to_symbol, _symbols
if add_blank:
sequence = list(filter(lambda x: x != len(_symbols), sequence))
if tp:
_symbols, _ = make_symbols(**tp)
_id_to_symbol = {i: s for i, s in enumerate(_symbols)}
result = ''
for symbol_id in sequence:
if symbol_id in _id_to_symbol:
s = _id_to_symbol[symbol_id]
# Enclose ARPAbet back in curly braces:
if len(s) > 1 and s[0] == '@':
s = '{%s}' % s[1:]
result += s
return result.replace('}{', ' ')
def text_to_sequence(text, cleaner_names, tp=None, add_blank=False):
'''Converts a string of text to a sequence of IDs corresponding to the symbols in the text.
The text can optionally have ARPAbet sequences enclosed in curly braces embedded
in it. For example, "Turn left on {HH AW1 S S T AH0 N} Street."
Args:
text: string to convert to a sequence
cleaner_names: names of the cleaner functions to run the text through
Returns:
List of integers corresponding to the symbols in the text
'''
# pylint: disable=global-statement
global _symbol_to_id, _symbols
if tp:
_symbols, _ = make_symbols(**tp)
_symbol_to_id = {s: i for i, s in enumerate(_symbols)}
sequence = []
# Check for curly braces and treat their contents as ARPAbet:
while text:
m = _CURLY_RE.match(text)
if not m:
sequence += _symbols_to_sequence(_clean_text(text, cleaner_names))
break
sequence += _symbols_to_sequence(_clean_text(m.group(1),
cleaner_names))
sequence += _arpabet_to_sequence(m.group(2))
text = m.group(3)
if add_blank:
sequence = intersperse(
sequence, len(_symbols)
) # add a blank token (new), whose id number is len(_symbols)
return sequence
def main(args): # pylint: disable=redefined-outer-name
# pylint: disable=global-variable-undefined
global meta_data_train, meta_data_eval, symbols, phonemes, speaker_mapping
# Audio processor
ap = AudioProcessor(**c.audio)
if 'characters' in c.keys():
symbols, phonemes = make_symbols(**c.characters)
# DISTRUBUTED
if num_gpus > 1:
init_distributed(args.rank, num_gpus, args.group_id,
c.distributed["backend"], c.distributed["url"])
num_chars = len(phonemes) if c.use_phonemes else len(symbols)
# load data instances
meta_data_train, meta_data_eval = load_meta_data(c.datasets)
# set the portion of the data used for training
if 'train_portion' in c.keys():
meta_data_train = meta_data_train[:int(len(meta_data_train) * c.train_portion)]
if 'eval_portion' in c.keys():
meta_data_eval = meta_data_eval[:int(len(meta_data_eval) * c.eval_portion)]
# parse speakers
num_speakers, speaker_embedding_dim, speaker_mapping = parse_speakers(c, args, meta_data_train, OUT_PATH)
model = setup_model(num_chars, num_speakers, c, speaker_embedding_dim)
# scalers for mixed precision training
scaler = torch.cuda.amp.GradScaler() if c.mixed_precision else None
scaler_st = torch.cuda.amp.GradScaler() if c.mixed_precision and c.separate_stopnet else None
params = set_weight_decay(model, c.wd)
optimizer = RAdam(params, lr=c.lr, weight_decay=0)
if c.stopnet and c.separate_stopnet:
optimizer_st = RAdam(model.decoder.stopnet.parameters(),
lr=c.lr,
weight_decay=0)
else:
optimizer_st = None
# setup criterion
criterion = TacotronLoss(c, stopnet_pos_weight=10.0, ga_sigma=0.4)
if args.restore_path:
checkpoint = torch.load(args.restore_path, map_location='cpu')
try:
print(" > Restoring Model.")
model.load_state_dict(checkpoint['model'])
# optimizer restore
print(" > Restoring Optimizer.")
optimizer.load_state_dict(checkpoint['optimizer'])
if "scaler" in checkpoint and c.mixed_precision:
print(" > Restoring AMP Scaler...")
scaler.load_state_dict(checkpoint["scaler"])
if c.reinit_layers:
raise RuntimeError
except (KeyError, RuntimeError):
print(" > Partial model initialization.")
model_dict = model.state_dict()
model_dict = set_init_dict(model_dict, checkpoint['model'], c)
# torch.save(model_dict, os.path.join(OUT_PATH, 'state_dict.pt'))
# print("State Dict saved for debug in: ", os.path.join(OUT_PATH, 'state_dict.pt'))
model.load_state_dict(model_dict)
del model_dict
for group in optimizer.param_groups:
group['lr'] = c.lr
print(" > Model restored from step %d" % checkpoint['step'],
flush=True)
args.restore_step = checkpoint['step']
else:
args.restore_step = 0
if use_cuda:
model.cuda()
criterion.cuda()
# DISTRUBUTED
if num_gpus > 1:
model = apply_gradient_allreduce(model)
if c.noam_schedule:
scheduler = NoamLR(optimizer,
warmup_steps=c.warmup_steps,
last_epoch=args.restore_step - 1)
else:
scheduler = None
num_params = count_parameters(model)
print("\n > Model has {} parameters".format(num_params), flush=True)
if 'best_loss' not in locals():
best_loss = float('inf')
# define data loaders
train_loader = setup_loader(ap,
model.decoder.r,
is_val=False,
verbose=True)
eval_loader = setup_loader(ap, model.decoder.r, is_val=True)
global_step = args.restore_step
for epoch in range(0, c.epochs):
c_logger.print_epoch_start(epoch, c.epochs)
# set gradual training
if c.gradual_training is not None:
r, c.batch_size = gradual_training_scheduler(global_step, c)
c.r = r
model.decoder.set_r(r)
if c.bidirectional_decoder:
model.decoder_backward.set_r(r)
train_loader.dataset.outputs_per_step = r
eval_loader.dataset.outputs_per_step = r
train_loader = setup_loader(ap,
model.decoder.r,
is_val=False,
dataset=train_loader.dataset)
eval_loader = setup_loader(ap,
model.decoder.r,
is_val=True,
dataset=eval_loader.dataset)
print("\n > Number of output frames:", model.decoder.r)
# train one epoch
train_avg_loss_dict, global_step = train(train_loader, model,
criterion, optimizer,
optimizer_st, scheduler, ap,
global_step, epoch, scaler,
scaler_st)
# eval one epoch
eval_avg_loss_dict = evaluate(eval_loader, model, criterion, ap,
global_step, epoch)
c_logger.print_epoch_end(epoch, eval_avg_loss_dict)
target_loss = train_avg_loss_dict['avg_postnet_loss']
if c.run_eval:
target_loss = eval_avg_loss_dict['avg_postnet_loss']
best_loss = save_best_model(
target_loss,
best_loss,
model,
optimizer,
global_step,
epoch,
c.r,
OUT_PATH,
scaler=scaler.state_dict() if c.mixed_precision else None)