Update modeling_internvl_chat.py

modeling_internvl_chat.py

@@ -325,6 +325,9 @@ class InternVLChatModel(PreTrainedModel):
                 vit_embeds = visual_features
             else:
                 vit_embeds = self.extract_feature(pixel_values)
+
+            print(vit_embeds.shape)
+                
             input_embeds = self.language_model.get_input_embeddings()(input_ids)
             B, N, C = input_embeds.shape
             input_embeds = input_embeds.reshape(B * N, C)