init

llava/serve/gradio_web_server.py

@@ -31,6 +31,7 @@ from PIL import Image
 import io
 import base64
 
+
 def resize_image(image, default_width=(1344, 896), request_width=None):
     # 如果 request 中指定了 width，则使用传入的值
     if request_width:
@@ -38,7 +39,7 @@ def resize_image(image, default_width=(1344, 896), request_width=None):
 
     original_width, original_height = image.size
 
-    print(original_width,original_height)
+    print("Original size:", original_width, original_height)
 
     # 根据宽高比决定 resize 逻辑
     if original_width >= original_height:
@@ -54,7 +55,14 @@ def resize_image(image, default_width=(1344, 896), request_width=None):
 
     # 调整图像大小
     resized_image = image.resize((new_width, new_height))
-    print(new_width,new_height)
+    print("After initial resize:", new_width, new_height)
+
+    # 如果高度仍然超过 2016，则将图片固定调整为 896x2016
+    if new_height > 2016:
+        new_width, new_height = 672, 2016
+        resized_image = resized_image.resize((new_width, new_height))
+        print("Adjusted to fixed size:", new_width, new_height)
+
     return resized_image
 
 

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "uground_demo_test"
-version = "3.5"
+version = "3.6"
 description = "Navigating the Digital World as Humans Do: Universal Visual Grounding for GUI Agents"
 readme = "README.md"
 requires-python = ">=3.8"