Refactor vctk unittest (#1134)

test/torchaudio_unittest/datasets/vctk_test.py

@@ -12,7 +12,7 @@ from torchaudio_unittest.common_utils import (
 )
 
 # Used to generate a unique utterance for each dummy audio file
-UTTERANCE = [
+_UTTERANCE = [
     'Please call Stella',
     'Ask her to bring these things',
     'with her from the store',
@@ -27,16 +27,12 @@ UTTERANCE = [
 ]
 
 
-class TestVCTK(TempDirMixin, TorchaudioTestCase):
-    backend = 'default'
-
-    root_dir = None
-    samples = []
-
-    @classmethod
-    def setUpClass(cls):
-        cls.root_dir = cls.get_base_temp_dir()
-        dataset_dir = os.path.join(cls.root_dir, 'VCTK-Corpus-0.92')
+def get_mock_dataset(root_dir):
+    """
+    root_dir: root directory of the mocked data
+    """
+    mocked_samples = []
+    dataset_dir = os.path.join(root_dir, 'VCTK-Corpus-0.92')
     os.makedirs(dataset_dir, exist_ok=True)
     sample_rate = 48000
     seed = 0
@@ -63,7 +59,7 @@ class TestVCTK(TempDirMixin, TorchaudioTestCase):
             save_wav(audio_file_path, data, sample_rate)
 
             txt_file_path = os.path.join(file_dir, filename[:-5] + '.txt')
-                utterance = UTTERANCE[utterance_id - 1]
+            utterance = _UTTERANCE[utterance_id - 1]
             with open(txt_file_path, 'w') as f:
                 f.write(utterance)
 
@@ -74,9 +70,21 @@ class TestVCTK(TempDirMixin, TorchaudioTestCase):
                 speaker_id,
                 utterance_id
             )
-                cls.samples.append(sample)
-
+            mocked_samples.append(sample)
             seed += 1
+    return mocked_samples
+
+
+class TestVCTK(TempDirMixin, TorchaudioTestCase):
+    backend = 'default'
+
+    root_dir = None
+    samples = []
+
+    @classmethod
+    def setUpClass(cls):
+        cls.root_dir = cls.get_base_temp_dir()
+        cls.samples = get_mock_dataset(cls.root_dir)
 
     def _test_vctk(self, dataset):
         num_samples = 0