Update model_class.py

model_class.py

@@ -26,6 +26,7 @@ class WhisperForAudioCaptioning(transformers.WhisperForConditionalGeneration):
         output_hidden_states: Optional[bool] = None,
         return_dict: Optional[bool] = None,
         forced_ac_decoder_ids: Optional[torch.LongTensor] = None, # added to be ignored when passed from trainer
+        decoder_position_ids: Optional[torch.LongTensor] = None,
     ) -> Union[Tuple[torch.Tensor], Seq2SeqLMOutput]:
         return super().forward(
             input_features=input_features,
@@ -43,6 +44,7 @@ class WhisperForAudioCaptioning(transformers.WhisperForConditionalGeneration):
             output_attentions=output_attentions,
             output_hidden_states=output_hidden_states,
             return_dict=return_dict,
+            decoder_position_ids=decoder_position_ids,
         )
     
     # copy-pasted and adapted from transformers.WhisperForConditionalGeneration.generate
@@ -156,3 +158,4 @@ class WhisperForAudioCaptioning(transformers.WhisperForConditionalGeneration):
             decoder_input_ids=decoder_input_ids,
             **kwargs,
         )
+