Update modeling_minicpmv.py

modeling_minicpmv.py

@@ -111,6 +111,7 @@ class MiniCPMV(MiniCPMVPreTrainedModel):
                     vision_embedding = self.resampler(vision_embedding, tgt_sizes)
                 else:
                     # get vision_embedding foreach
+                    print("HERE, NOT BATCH")
                     vision_embedding = []
                     for single_tgt_size, single_pixel_values in zip(tgt_sizes, all_pixel_values):
                         single_pixel_values = single_pixel_values.unsqueeze(0)
@@ -262,6 +263,7 @@ class MiniCPMV(MiniCPMVPreTrainedModel):
             else:
                 model_inputs["vision_hidden_states"] = vision_hidden_states
     
+            print(model_inputs)
             (
                 input_embeds,
                 vision_hidden_states,