strore only path to audio (don't read bytes)

Audio feature automatically decode local audio files, there is no need to open it and store bytes, you can just provide path to local file :)

cvss.py

@@ -134,12 +134,8 @@ class CVSS(datasets.GeneratorBasedBuilder):
           cols = line.rstrip().split('\t')
           assert len(cols) == 2, cols
           key, text = cols
-
-          audio_path = os.path.join(path, split, f'{key}.wav')
-          with open(audio_path, 'rb') as audio_f:
-            audio_bytes = audio_f.read()
-
-          audio = {'path': audio_path, 'bytes': audio_bytes}
+          audio = {'path': audio_path}
+          
           yield key, {
               'id': key,
               'text': text,