Update app.py

app.py

@@ -32,14 +32,19 @@ async def process_image_stream(image_path, prompt, max_tokens=512):
                 "max_tokens": max_tokens
             }))
             
-            # Initialize response
+            # Initialize response and token counter
             response = ""
+            token_count = 0
             
             # Receive streaming response
             async for message in websocket:
                 try:
                     data = json.loads(message)
                     if data["status"] == "generating":
+                        # Skip first three tokens if they match specific patterns
+                        if token_count < 3 and data["token"] in [" ", " \n", "\n", "<|im_start|>", "assistant"]:
+                            token_count += 1
+                            continue
                         response += data["token"]
                         yield response
                     elif data["status"] == "complete":
@@ -77,7 +82,9 @@ demo = gr.Interface(
     Upload an image and ask questions about it. The model will analyze the image and provide detailed answers to your queries.
     """,
     examples=[
-        ["example_images/example_1.jpg", "Describe this image", 128],
+        ["example_images/example_1.jpg", "What kind of cat is this?", 128],
+        ["example_images/example_2.jpg", "What color is this dress? ", 128],
+        ["example_images/example_3.jpg", "What is this image about?", 128],
     ]
 )