Update audioset.py

audioset.py

@@ -166,19 +166,19 @@ class AudioSet(datasets.GeneratorBasedBuilder):
             fileid = Path(wav_path).name
             sound = fileid2labels.get(fileid)
             duration = librosa.get_duration(filename=wav_path)
-            print(duration)
+            if duration <= 0:
+                continue
             try:
-                metadata = torchaudio.info(wav_path)
+                # metadata = torchaudio.info(wav_path)
                 # duration = math.ceil(metadata.num_frames / metadata.sample_rate) # in seconds
-                if duration > 0:
-                    yield guid, {
-                        "id": str(guid),
-                        "file": wav_path, 
-                        "audio": wav_path, 
-                        "sound": sound, 
-                        "label": sound, 
-                        "duration": duration
-                    }
+                yield guid, {
+                    "id": str(guid),
+                    "file": wav_path, 
+                    "audio": wav_path, 
+                    "sound": sound, 
+                    "label": sound, 
+                    "duration": duration
+                }
             except:
                 continue