Update app.py

app.py

@@ -40,10 +40,12 @@ if audio_file is not None:
         inputs = feature_extractor(clip.squeeze().numpy(), return_tensors="pt")
         input_features = inputs.input_features
         
-        # Pad input features to length 3000
+        # Ensure input features have the required length of 3000
         if input_features.shape[-1] < 3000:
             pad_length = 3000 - input_features.shape[-1]
-            input_features = torch.nn.functional.pad(input_features, (0, pad_length))
+            input_features = torch.nn.functional.pad(input_features, (0, pad_length), mode='constant', value=0)
+        elif input_features.shape[-1] > 3000:
+            input_features = input_features[:, :, :3000]
 
         # Measure inference time
         start_time = time.time()