def prepare_data(self):
# setup training set
if "timit" in self.hp.data:
train, val, test = TrainTestDataset.get_datasets(
path=self.hp.timit_path)
elif "buckeye" in self.hp.data:
train, val, test = TrainValTestDataset.get_datasets(
path=self.hp.buckeye_path, percent=self.hp.buckeye_percent)
else:
raise Exception("no such training data!")
if "libri" in self.hp.data:
libri_train = LibriSpeechDataset(path=self.hp.libri_path,
subset=self.hp.libri_subset,
percent=self.hp.libri_percent)
train = ConcatDataset([train, libri_train])
train.path = "\n\t+".join(
[dataset.path for dataset in train.datasets])
print(f"added libri ({len(libri_train)} examples)")
self.train_dataset = train
self.valid_dataset = val
self.test_dataset = test
line()
print("DATA:")
print(f"train: {self.train_dataset.path} ({len(self.train_dataset)})")
print(f"valid: {self.valid_dataset.path} ({len(self.valid_dataset)})")
print(f"test: {self.test_dataset.path} ({len(self.test_dataset)})")
line()
