target_dir.mkdir(exist_ok=True)
train_target_dir.mkdir(exist_ok=True)
val_target_dir.mkdir(exist_ok=True)
train_predictions_dir.mkdir(exist_ok=True)
val_predictions_dir.mkdir(exist_ok=True)
config_manager.dump_config()
script_batch_size = 5 * config['batch_size']
val_has_files = len([
batch_file for batch_file in val_predictions_dir.iterdir()
if batch_file.suffix == '.npy'
])
train_has_files = len([
batch_file for batch_file in train_predictions_dir.iterdir()
if batch_file.suffix == '.npy'
])
model = config_manager.load_model()
running_predictions = args.recompute_pred or (val_has_files
== 0) or (train_has_files == 0)
if running_predictions:
if args.store_predictions:
print(
'\nWARNING: storing predictions can take a lot of disk space (~40GB)\n'
)
train_meta = config_manager.train_datadir / 'train_metafile.txt'
test_meta = config_manager.train_datadir / 'test_metafile.txt'
train_samples, _ = load_files(
metafile=str(train_meta),
meldir=str(meldir),
num_samples=config['n_samples']) # (phonemes, mel)
val_samples, _ = load_files(
metafile=str(test_meta),
clear_logs=args.clear_logs,
clear_weights=args.clear_weights)
config_manager.dump_config()
config_manager.print_config()
train_samples, _ = load_files(metafile=str(config_manager.train_datadir / 'train_metafile.txt'),
meldir=str(config_manager.train_datadir / 'mels'),
num_samples=config['n_samples']) # (phonemes, mel)
val_samples, _ = load_files(metafile=str(config_manager.train_datadir / 'test_metafile.txt'),
meldir=str(config_manager.train_datadir / 'mels'),
num_samples=config['n_samples']) # (phonemes, text, mel)
# get model, prepare data for model, create datasets
pretrained = True
if pretrained:
model = config_manager.load_model(str(config_manager.weights_dir /'ckpt-90'))
else:
model = config_manager.get_model()
config_manager.compile_model(model)
data_prep = DataPrepper(config=config,
tokenizer=model.text_pipeline.tokenizer)
test_list = [data_prep(s) for s in val_samples]
train_dataset = Dataset(samples=train_samples,
preprocessor=data_prep,
batch_size=config['batch_size'],
mel_channels=config['mel_channels'],
shuffle=True)
val_dataset = Dataset(samples=val_samples,
preprocessor=data_prep,