Update ultravox_processing.py

ultravox_processing.py

@@ -157,7 +157,7 @@ class UltravoxProcessor(transformers.ProcessorMixin):
                 data["audio_values"] = x.input_features
             else:
                 data["audio_values"] = x.input_values
-            data["audio_len"] = x.attention_mask.sum(-1) - 1
+            data["audio_len"] = np.asarray(x.attention_mask).sum(-1) - 1
 
         if text is not None:
             #assert isinstance(