WIP

Author: Ye Shaokai

llava/model/language_model/llava_qwen.py

@@ -92,7 +92,6 @@ def forward(
         dpo_forward: Optional[bool] = False,
         cache_position=None,
     ) -> Union[Tuple, CausalLMOutputWithPast]:
-        print ('actions', actions)
         if inputs_embeds is None:
             (input_ids, position_ids, attention_mask, past_key_values, inputs_embeds, labels, action_idx) = self.prepare_inputs_labels_for_multimodal(input_ids, position_ids, attention_mask, past_key_values, labels, images, modalities, image_sizes)