checkpoint_path = args["<checkpoint>"]
text_list_file_path = args["<text_list_file>"]
dst_dir = args["<dst_dir>"]
max_decoder_steps = int(args["--max-decoder-steps"])
file_name_suffix = args["--file-name-suffix"]
checkpoint = torch.load(checkpoint_path)
checkpoints_dir = os.path.dirname(checkpoint_path)
with open(checkpoints_dir + '/ids_phones.json') as f:
phids = json.load(f)
model = Tacotron(
n_vocab=len(phids) + 1,
embedding_dim=256,
mel_dim=hparams.num_mels,
linear_dim=hparams.num_freq,
r=hparams.outputs_per_step,
padding_idx=hparams.padding_idx,
use_memory_mask=hparams.use_memory_mask,
)
checkpoint = torch.load(checkpoint_path)
checkpoints_dir = os.path.dirname(checkpoint_path)
with open(checkpoints_dir + '/ids_phones.json') as f:
phids = json.load(f)
phids = dict(phids)
model.load_state_dict(checkpoint["state_dict"])
model.decoder.max_decoder_steps = max_decoder_steps
os.makedirs(dst_dir, exist_ok=True)
collate_fn=collate_fn, pin_memory=hparams.pin_memory)
exp_name = os.path.basename(exp_dir)
if use_assistant:
assistant = RemoteTracker(exp_name, projects_dir='Arabic',
upload_source_files=[os.getcwd() + '/' + 'conf/arabic.conf',
os.getcwd() + '/local/model.py',
os.getcwd() + '/local/train_phones.py'])
else:
assistant = None
# Model
model = Tacotron(n_vocab=1+ len(ph_ids),
embedding_dim=256,
mel_dim=hparams.num_mels,
linear_dim=hparams.num_freq,
r=hparams.outputs_per_step,
padding_idx=hparams.padding_idx,
use_memory_mask=hparams.use_memory_mask,
)
model = model.cuda()
#model = DataParallelFix(model)
optimizer = optim.Adam(model.parameters(),
lr=hparams.initial_learning_rate, betas=(
hparams.adam_beta1, hparams.adam_beta2),
weight_decay=hparams.weight_decay)
# Load checkpoint
if checkpoint_path:
print("Load checkpoint from: {}".format(checkpoint_path))
checkpoint = torch.load(checkpoint_path)
file_name_suffix = args["--file-name-suffix"]
checkpoint_acousticmodel = torch.load(checkpoint_path_acousticmodel)
checkpoint_vocoder = torch.load(checkpoint_path_vocoder)
checkpoints_dir = os.path.dirname(checkpoint_path_acousticmodel)
with open(checkpoints_dir + '/ids_phones.json') as f:
phids = json.load(f)
phids = dict(phids)
acousticmodel = Tacotron(
n_vocab=len(phids) + 1,
embedding_dim=256,
mel_dim=hparams.num_mels,
linear_dim=hparams.num_freq,
r=hparams.outputs_per_step,
padding_idx=hparams.padding_idx,
use_memory_mask=hparams.use_memory_mask,
)
vocoder_model = WaveLSTM(
n_vocab=257,
embedding_dim=256,
mel_dim=hparams.num_mels,
linear_dim=hparams.num_freq,
r=hparams.outputs_per_step,
padding_idx=hparams.padding_idx,
use_memory_mask=hparams.use_memory_mask,
)
