def _iter(self):
for name, path in self.dataset_dir.items():
if name == 'lapsbm':
continue
try:
dataset_cls = get_from_module('datasets*', name, regex=True)
dataset = dataset_cls(dataset_dir=path)
for d in dataset._iter():
yield {
'duration': d['duration'],
'input': d['input'],
'label': d['label'],
'speaker': '%s_%s' % (str(dataset), d['speaker']),
'dataset': 'train'
}
except ValueError as e:
self._logger.warning('Skipping dataset %s: %s' % (name, e))
# Test and valid set
lapsbm = LapsBM(dataset_dir=self.dataset_dir['lapsbm'], split=True)
for d in lapsbm._iter():
yield {
'duration': d['duration'],
'input': d['input'],
'label': d['label'],
'speaker': '%s_%s' % (str(dataset), d['speaker']),
'dataset': d['dataset']
}
def _iter(self):
for name, path in self.dataset_dir.items():
if name == 'cv_corpus_v1':
continue
try:
dataset_cls = get_from_module('datasets*', name, regex=True)
dataset = dataset_cls(dataset_dir=path)
print(dataset)
for d in dataset._iter():
yield {
'duration': d['duration'],
'input': d['input'],
'label': d['label'],
'speaker': '%s_%s' % (str(dataset), d['speaker']),
'dataset': 'train'
}
except ValueError as e:
self._logger.warning('Skipping dataset %s: %s' % (name, e))
# Test and valid set
cvc = CVC(dataset_dir=self.dataset_dir['cv_corpus_v1'])
for d in cvc._iter():
yield {
'duration': d['duration'],
'input': d['input'],
'label': d['label'],
'audio_file': d['audio_file'],
'dataset': d['dataset']
}
if args.dataset and args.file:
print('Both dataset and file args was set. Ignoring file args.')
# GPU configuration
setup_gpu(args.gpu, args.allow_growth)
# Loading model
model, meta = load_model(args.model, return_meta=True,
mode='predict', decoder=(not args.no_decoder))
args = HParams(**meta['training_args']).update(vars(args_nondefault))
# Features extractor
input_parser = utils.get_from_module('preprocessing.audio',
args.input_parser,
params=args.input_parser_params)
# Recovering text parser
label_parser = utils.get_from_module('preprocessing.text',
args.label_parser,
params=args.label_parser_params)
if args.dataset is not None:
data_gen = DatasetGenerator(input_parser, label_parser,
batch_size=1, seed=0, mode='predict',
shuffle=False)
test_flow = data_gen.flow_from_fname(args.dataset,
datasets=args.subset)
else:
test_flow = DatasetIterator(np.array([args.file]), None,
def get(identifier):
return get_from_module(identifier, globals(), 'objective')
def get(identifier):
return get_from_module(identifier, globals(), 'initialization')
def get(identifier):
return get_from_module(identifier, globals(), 'optimizer', instantiate=True)
def get(identifier):
return get_from_module(identifier, globals(), 'activation function')
parser.add_argument('--parser', type=str, required=True)
parser.add_argument('--parser_params', nargs='+', default=[])
parser.add_argument('--output_file', type=str, default=None)
parser.add_argument('--input_parser', type=str, default=None)
parser.add_argument('--input_parser_params', nargs='+', default=[])
parser.add_argument('--label_parser', type=str, default=None)
parser.add_argument('--label_parser_params', nargs='+', default=[])
parser.add_argument('--override', action='store_true')
args = parser.parse_args()
parser = utils.get_from_module('datasets*', args.parser, regex=True)
print('input args ', args.input_parser_params)
input_parser = utils.get_from_module('preprocessing.audio',
args.input_parser,
params=args.input_parser_params)
label_parser = utils.get_from_module('preprocessing.text',
args.label_parser,
params=args.label_parser_params)
dataset = parser(args.dataset_dir,
**HParams().parse(args.parser_params).values())
output_file = dataset.to_h5(fname=args.output_file,
input_parser=input_parser,
label_parser=label_parser,
override=args.override)
def get(identifier):
return get_from_module(identifier, globals(), 'activation function')
model, meta = load_model(args.load, return_meta=True)
logger.info('Loading parameters...')
args = HParams(**meta['training_args']).update(vars(args_nondefault))
epoch_offset = len(meta['epochs'])
logger.info('Current epoch: %d' % epoch_offset)
if args_nondefault.lr:
logger.info('Setting current learning rate to %f...' % args.lr)
K.set_value(model.optimizer.lr, args.lr)
else:
logger.info('Creating model...')
# Recovering all valid models
model_fn = utils.get_from_module('core.models', args.model)
# Loading model
model = model_fn(**(HParams().parse(args.model_params).values()))
logger.info('Setting the optimizer...')
# Optimization
if args.opt.strip().lower() == 'sgd':
opt = SGD(lr=args.lr,
momentum=args.momentum,
clipnorm=args.clipnorm)
elif args.opt.strip().lower() == 'adam':
opt = Adam(lr=args.lr, clipnorm=args.clipnorm)
# Compile with dummy loss
model.compile(loss={
'ctc': ctc_dummy_loss,
for audio_fname in args.source:
with sr.AudioFile(audio_fname) as source:
audios.append((r.record(source), audio_fname))
# read the entire audio file
if args.model is not None:
setup_gpu(args.gpu, args.allow_growth)
model, meta = utils.load_model(args.model,
return_meta=True,
mode='predict')
training_args = meta['training_args']
# Features extractor
input_parser = utils.get_from_module('preprocessing.audio',
training_args['feats'],
params=training_args['feats_params'])
# Recovering text parser
label_parser = utils.get_from_module('preprocessing.text',
training_args['label_parser'],
params=training_args['label_parser_params']
)
data_it = DatasetIterator(np.array([f for a, f in audios]),
input_parser=input_parser,
label_parser=label_parser)
model_predictions = model.predict_generator(
data_it, val_samples=len(audios))
def get(identifier):
return get_from_module(identifier,
globals(),
'optimizer',
instantiate=True)
def get(identifier):
return get_from_module(identifier, globals(), 'initialization')
def get(identifier, kwargs=None):
return get_from_module(identifier, globals(), 'optimizer', instantiate=True, kwargs=kwargs)
if __name__ == '__main__':
parser = argparse.ArgumentParser(
description=
'Generates a preprocessed dataset (hdf5 file) by providing the path to the dataset and the correct parser.'
)
parser.add_argument('--dataset_dir', type=str, default=None)
parser.add_argument('--parser', type=str, required=True)
parser.add_argument('--parser_params', nargs='+', default=[])
parser.add_argument('--output_file', type=str, default=None)
parser.add_argument('--override', action='store_true')
args = parser.parse_args()
parser = utils.get_from_module('datasets*', args.parser, regex=True)
input_parser = preprocessing.MFCC()
label_parser = preprocessing.SimpleCharParser()
dataset = parser(args.dataset_dir,
**HParams().parse(args.parser_params).values())
output_file = dataset.to_h5(fname=args.output_file,
input_parser=input_parser,
label_parser=label_parser,
override=args.override)
print('Dataset %s saved at %s' % (parser.name, output_file))
