Remove empty files in the Vocal Sound audio scenario (#3313)

src/helm/benchmark/scenarios/audio_language/vocal_sound_scenario.py

@@ -15,6 +15,7 @@
 from tqdm import tqdm
 from helm.common.media_object import MediaObject, MultimediaObject
 from helm.common.general import ensure_file_downloaded
+from helm.common.audio_utils import remove_audio_clips_with_zero_frames
 
 
 class VocalSoundScenario(Scenario):
@@ -49,10 +50,9 @@ def get_instances(self, output_path: str) -> List[Instance]:
         down_loading_path = os.path.join(output_path, "download")
         ensure_file_downloaded(VocalSoundScenario.DOWNLOADING_URL, down_loading_path, unpack=True)
         wav_save_dir = os.path.join(down_loading_path, "audio_16k")
+        remove_audio_clips_with_zero_frames(wav_save_dir)
         for file_name in tqdm(os.listdir(wav_save_dir)):
             local_audio_path = os.path.join(wav_save_dir, file_name)
-            assert os.path.exists(local_audio_path), f"Audio file does not exist at path: {local_audio_path}"
-
             answer = file_name.split("_")[-1].split(".")[0]
             input = Input(
                 multimedia_content=MultimediaObject([MediaObject(content_type="audio/wav", location=local_audio_path)])

src/helm/common/audio_utils.py

@@ -62,3 +62,17 @@ def use_ffmpeg_to_extract_audio_from_video(input_video_path: str, output_audio_p
         subprocess.run(["ffmpeg", "-i", input_video_path, "-q:a", "0", "-map", "a", output_audio_path], check=True)
     except (subprocess.CalledProcessError, FileNotFoundError):
         raise ValueError("Please install ffmpeg using `bash install-shelm-extras.sh` first to extract audio files.")
+
+
+def remove_audio_clips_with_zero_frames(audio_directory: str) -> None:
+    # Iterate through all files in the directory
+    for filename in os.listdir(audio_directory):
+        if filename.endswith(".wav"):
+            file_path = os.path.join(audio_directory, filename)
+            try:
+                with sf.SoundFile(file_path) as audio_file:
+                    if len(audio_file) == 0:  # Check if the file has zero frames
+                        print(f"Removing empty file: {filename}")
+                        os.remove(file_path)
+            except RuntimeError:
+                print(f"Skipping invalid audio file: {filename}")