Update app.py

app.py

@@ -320,6 +320,8 @@ def add_message(history, message):
     #     return history, gr.MultimodalTextbox(value=None, interactive=False)
     # else:
     for x in message["files"]:
+        if "realcase_video.jpg" in x:
+            x = x.replace("realcase_video.jpg", "realcase_video.mp4")
         history.append(((x,), None))
     if message["text"] is not None:
         history.append((message["text"], None))
@@ -408,6 +410,7 @@ def bot(history, temperature, top_p, max_output_tokens):
                     with open(file_path, "rb") as src, open(filename, "wb") as dst:
                         dst.write(src.read())
     
+    image_tensor = []
     if not is_video:
         image_tensor = [
             our_chatbot.image_processor.preprocess(f, return_tensors="pt")["pixel_values"][
@@ -418,14 +421,14 @@ def bot(history, temperature, top_p, max_output_tokens):
             for f in image_list
         ]
         image_tensor = torch.stack(image_tensor)
-    else:
+    elif num_new_images > 0:
         image_tensor = our_chatbot.image_processor.preprocess(image_list, return_tensors="pt")["pixel_values"].half().to(our_chatbot.model.device)
-
-
-    image_token = DEFAULT_IMAGE_TOKEN * num_new_images if not is_video else DEFAULT_IMAGE_TOKEN * num_new_images
+    
+    
+    image_token = DEFAULT_IMAGE_TOKEN * num_new_images + "\n"
     
     inp = text
-    inp = image_token + "\n" + inp
+    inp = image_token + inp
     our_chatbot.conversation.append_message(our_chatbot.conversation.roles[0], inp)
     # image = None
     our_chatbot.conversation.append_message(our_chatbot.conversation.roles[1], None)
@@ -450,14 +453,20 @@ def bot(history, temperature, top_p, max_output_tokens):
     )
     print(our_chatbot.model.device)
     print(input_ids.device)
-    print(image_tensor.device)
+    # print(image_tensor.device)
     
     
+    if is_video:
+        input_image_tensor = [image_tensor]
+    elif num_new_images > 0:
+        input_image_tensor = image_tensor
+    else:
+        input_image_tensor = None
     
     generate_kwargs = dict(
         inputs=input_ids,
         streamer=streamer,
-        images=[image_tensor] if is_video else image_tensor,
+        images=input_image_tensor,
         do_sample=True,
         temperature=temperature,
         top_p=top_p,
@@ -610,7 +619,7 @@ with gr.Blocks(
                             "files": [
                                 f"{cur_dir}/examples/realcase_math.jpg",
                             ],
-                            "text": "Find the measure of angle 3.",
+                            "text": "Find the measure of angle 3. Please provide a step by step solution.",
                         }
                     ],
                     [