Upload app.py with huggingface_hub

app.py

@@ -112,16 +112,26 @@ def process_sample(sample, language, text_col="sentence"):
         if audio is None:
             return None, "No audio"
 
-        # Handle different audio formats
+        # Handle different audio formats (dict, AudioDecoder, etc.)
         audio_array = None
         sample_rate = 16000
 
         if isinstance(audio, dict):
             audio_array = audio.get("array")
             sample_rate = audio.get("sampling_rate", 16000)
+        elif hasattr(audio, 'array') and hasattr(audio, 'sampling_rate'):
+            # AudioDecoder or similar object with attributes
+            audio_array = audio.array
+            sample_rate = audio.sampling_rate
         elif hasattr(audio, '__getitem__'):
-            audio_array = audio.get("array") if hasattr(audio, 'get') else audio["array"]
-            sample_rate = audio.get("sampling_rate", 16000) if hasattr(audio, 'get') else audio.get("sampling_rate", 16000)
+            # Dict-like access
+            try:
+                audio_array = audio["array"]
+                sample_rate = audio.get("sampling_rate", 16000) if hasattr(audio, 'get') else 16000
+            except (KeyError, TypeError):
+                return None, f"Cannot access audio array: {type(audio)}"
+        else:
+            return None, f"Unknown audio format: {type(audio)}"
 
         if audio_array is None or len(audio_array) == 0:
             return None, "Empty audio"