Update ultravox_processing.py

ultravox_processing.py

@@ -141,7 +141,7 @@ class UltravoxProcessor(transformers.ProcessorMixin):
                 assert sampling_rate is not None, "Sampling rate must be provided."
                 audio_len = 30 * sampling_rate
             else:
-                audio_len = audio.shape[-1]
+                audio_len = max([a.shape[-1] for a in audio])
             # It's guaranteed that the number of frames is less than or equal to this amount.
             # For Whisper this is exact AFAICT, but for Wav2Vec2 it's an upper bound.
             # Currently, StackAudioFrames makes sure an over-estimation won't cause issues by padding the audio embeddings.