Update app.py

app.py

@@ -35,28 +35,28 @@ tokenizer, llava_model, image_processor, context_len = load_pretrained_model(
 @spaces.GPU
 def bot_streaming(message, history):
     print(message)
-    image = None
+    image_path = None
     
     # Check if there's an image in the current message
     if message["files"]:
         # message["files"][-1] could be a dictionary or a string
         if isinstance(message["files"][-1], dict):
-            image = message["files"][-1]["path"]
+            image_path = message["files"][-1]["path"]
         else:
-            image = message["files"][-1]
+            image_path = message["files"][-1]
     else:
-        # If no image in the current message, look in the history for the last image
+        # If no image in the current message, look in the history for the last image path
         for hist in history:
             if isinstance(hist[0], tuple):
-                image = hist[0][0]
+                image_path = hist[0][0]
     
-    # Error handling if no image is found
-    if image is None:
+    # Error handling if no image path is found
+    if image_path is None:
         raise gr.Error("You need to upload an image for LLaVA to work.")
     
-    # Load the image if it's a path, otherwise use the existing PIL image
-    if isinstance(image, str):
-        image = Image.open(image).convert('RGB')
+    # If the image_path is a string, no need to load it into a PIL image
+    # Just use the path directly in the next steps
+    print(f"\033[91m{image_path}, {type(image_path)}\033[0m")
     
     # Generate the prompt for the model
     prompt = message['text']
@@ -105,7 +105,8 @@ with gr.Blocks(fill_height=True) as demo:
         fn=bot_streaming,
         title="FinLLaVA",
         examples=[{"text": "What is on the flower?", "files": ["./bee.jpg"]},
-                  {"text": "How to make this pastry?", "files": ["./baklava.png"]}],
+                  {"text": "How to make this pastry?", "files": ["./baklava.png"]},
+                  {"text":"What is this?","files":["http://images.cocodataset.org/val2017/000000039769.jpg"]}],
         stop_btn="Stop Generation",
         multimodal=True,
         textbox=chat_input,