Update README.md

README.md

@@ -10,7 +10,7 @@ Below is the code of a quick demo:
 
 ```
 # pip install "transformers>=4.42" pillow openai
-# 并启动你的 vLLM 服务，例如：
+# start vllm server like:
 # vllm serve osunlp/GUI-Drag-7B --tensor-parallel-size 1 --dtype bfloat16 --port 8000
 
 import base64
@@ -23,7 +23,7 @@ from PIL import Image
 from transformers.models.qwen2_vl.image_processing_qwen2_vl_fast import smart_resize as qwen_smart_resize
 
 MODEL_NAME = "GUI-Drag-7B"
-BASE_URL = "http://localhost:8000/v1"   # 替换成你的 vLLM 服务端口
+BASE_URL = "http://localhost:8000/v1" # replace it with your own port
 FN_CALL_TEMPLATE = """You are a helpful assistant.
 # Tools
 You may call one or more functions to assist with the user query.
@@ -61,17 +61,17 @@ def process_simple_drag_response(parsed_responses):
 def resize_back(coords, original_size, resized_size):
     ox, oy = original_size
     rx, ry = resized_size
-    return round(coords[0] * ox / rx), round(coords[1] * oy / ry)
+    return round(coords[0] * rx / ox), round(coords[1] * ry / oy)
 
 def demo():
     image_path = Path("demo_image.png")
-    instruction = "Drag to select the highlighted paragraph."
+    instruction = "Drag to select the first sentence of the first paragraph."
     image = Image.open(image_path)
 
     resized_h, resized_w = qwen_smart_resize(
         image.height, image.width,
-        max_pixels=2_116_800,
-        min_pixels=12_544,
+        max_pixels=2116800,
+        min_pixels=12544,
     )
 
     messages = [