dataset path. (#316)

Fix test for datasets.

test/assets/LibriSpeech/dev-clean/1272/128104/1272-128104.trans.txt

-1272-128104-0000 THIS IS A TEST.
+1272-128104-0000 MISTER QUILTER

test/test_datasets.py

@@ -7,9 +7,12 @@ from torchaudio.datasets.utils import DiskCache
 from torchaudio.datasets.vctk import VCTK
 from torchaudio.datasets.yesno import YESNO
 
+import common_utils
+
 
 class TestDatasets(unittest.TestCase):
-    path = "assets"
+    test_dirpath, test_dir = common_utils.create_temp_assets_dir()
+    path = os.path.join(test_dirpath, "assets")
 
     def test_yesno(self):
         data = YESNO(self.path, return_dict=True)

torchaudio/datasets/librispeech.py

 import os
 
-import torchaudio
 from torch.utils.data import Dataset
+
+import torchaudio
 from torchaudio.datasets.utils import (
     download_url,
     extract_archive,
@@ -19,7 +20,9 @@ def load_librispeech_item(fileid, path, ext_audio, ext_txt):
 
     file_text = speaker + "-" + chapter + ext_txt
     file_text = os.path.join(path, speaker, chapter, file_text)
-    file_audio = speaker + "-" + chapter + "-" + utterance + ext_audio
+
+    fileid_audio = speaker + "-" + chapter + "-" + utterance
+    file_audio = fileid_audio + ext_audio
     file_audio = os.path.join(path, speaker, chapter, file_audio)
 
     # Load audio
@@ -28,11 +31,11 @@ def load_librispeech_item(fileid, path, ext_audio, ext_txt):
     # Load text
     for line in open(file_text):
         fileid_text, content = line.strip().split(" ", 1)
-        if file_audio == fileid_text:
+        if fileid_audio == fileid_text:
             break
     else:
         # Translation not found
-        raise ValueError
+        raise FileNotFoundError("Translation not found for " + fileid_audio)
 
     return {
         "speaker_id": speaker,