debug

nnet/augmentation.py

@@ -543,14 +543,16 @@ def data_augmentation(speech,
         speech = (scale * speech + noise) / 2
 
     if "codec" in augmentations:
+        final_shape = speech.shape[1]
         configs = [
-            {"format": "wav", "encoding": 'ULAW', "bits_per_sample": 8},
-            {"format": "gsm"},
-            {"format": "mp3", "compression": -9},
-            {"format": "vorbis", "compression": -1}
+            ({"format": "wav", "encoding": 'ULAW', "bits_per_sample": 8}, "8 bit mu-law"),
+            ({"format": "gsm"}, "GSM-FR"),
+            ({"format": "mp3", "compression": -9}, "MP3"),
+            ({"format": "vorbis", "compression": -1}, "Vorbis")
         ]
         param, title = random.choice(configs)
         speech = torchaudio.functional.apply_codec(speech, sample_rate, **param)
+        speech = speech[:, :final_shape]
 
     return speech