diff --git a/scripts/user_scripts/tasks.py b/scripts/user_scripts/tasks.py
index 76d027a..bc858e5 100644
--- a/scripts/user_scripts/tasks.py
+++ b/scripts/user_scripts/tasks.py
@@ -20,20 +20,24 @@ log = logging.getLogger("lama-cleaner")
 
 def find_free_port() -> int:
     with closing(socket.socket(socket.AF_INET, socket.SOCK_STREAM)) as s:
-        s.bind(('', 0))
+        s.bind(("", 0))
         s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
         return s.getsockname()[1]
 
+
 CONFIG_PATH = "config.json"
 
+
 class MODEL(str, Enum):
     SD15 = "sd1.5"
-    LAMA = 'lama'
+    LAMA = "lama"
+
 
 class DEVICE(str, Enum):
     CUDA = "cuda"
     CPU = "cpu"
 
+
 @task
 def info(c):
     print("Environment information".center(60, "-"))
@@ -44,57 +48,83 @@ def info(c):
         c.run("python --version")
         c.run("which pip")
         c.run("pip --version")
-        c.run("pip list | grep lama")
+        c.run('pip list | grep "torch\|lama\|diffusers\|opencv\|cuda"')
     except:
         pass
-    print("-"*60)
+    print("-" * 60)
+
 
 @task(pre=[info])
 def config(c, disable_device_choice=False):
     # TODO: 提示选择模型，选择设备，端口，host
     # 如果是 sd 模型，提示接受条款和输入 huggingface token
-    model = Prompt.ask("Choice model", choices=[MODEL.SD15, MODEL.LAMA], default=MODEL.SD15)
+    model = Prompt.ask(
+        "Choice model", choices=[MODEL.SD15, MODEL.LAMA], default=MODEL.SD15
+    )
 
     hf_access_token = ""
     if model == MODEL.SD15:
         while True:
-            hf_access_token = Prompt.ask("Huggingface access token (https://huggingface.co/docs/hub/security-tokens)")
+            hf_access_token = Prompt.ask(
+                "Huggingface access token (https://huggingface.co/docs/hub/security-tokens)"
+            )
             if hf_access_token == "":
                 log.warning("Access token is required to download model")
             else:
-                break  
+                break
 
     if disable_device_choice:
         device = DEVICE.CPU
     else:
-        device = Prompt.ask("Choice device", choices=[DEVICE.CUDA, DEVICE.CPU], default=DEVICE.CUDA)
+        device = Prompt.ask(
+            "Choice device", choices=[DEVICE.CUDA, DEVICE.CPU], default=DEVICE.CUDA
+        )
         if device == DEVICE.CUDA:
             import torch
+
             if not torch.cuda.is_available():
-                log.warning("Did not find CUDA device on your computer, fallback to cpu")
+                log.warning(
+                    "Did not find CUDA device on your computer, fallback to cpu"
+                )
                 device = DEVICE.CPU
 
-    configs = {"model": model, "device": device, "hf_access_token": hf_access_token}
+    desktop = Confirm.ask("Start as desktop app?")
+
+    configs = {
+        "model": model,
+        "device": device,
+        "hf_access_token": hf_access_token,
+        "desktop": desktop,
+    }
     log.info(f"Save config to {CONFIG_PATH}")
-    with open(CONFIG_PATH, 'w', encoding='utf-8') as f:
+    with open(CONFIG_PATH, "w", encoding="utf-8") as f:
         json.dump(configs, f, indent=2, ensure_ascii=False)
     log.info(f"Config finish, you can close this window.")
 
 
-@task(pre=[info]) 
+@task(pre=[info])
 def start(c):
     if not os.path.exists(CONFIG_PATH):
         log.info("Config file not exists, please run config.sh first")
         exit()
 
     log.info(f"Load config from {CONFIG_PATH}")
-    with open(CONFIG_PATH, 'r', encoding='utf-8') as f:
+    with open(CONFIG_PATH, "r", encoding="utf-8") as f:
         configs = json.load(f)
 
-    model = configs['model']
-    device = configs['device']
-    hf_access_token = configs['hf_access_token']
+    model = configs["model"]
+    device = configs["device"]
+    hf_access_token = configs["hf_access_token"]
+    desktop = configs["desktop"]
     port = find_free_port()
     log.info(f"Using random port: {port}")
 
-    c.run(f"lama-cleaner --model {model} --device {device} --hf_access_token={hf_access_token} --port {port} --gui --gui-size 1400 900")
\ No newline at end of file
+    if desktop:
+        c.run(
+            f"lama-cleaner --model {model} --device {device} --hf_access_token={hf_access_token} --port {port} --gui --gui-size 1400 900"
+        )
+    else:
+        c.run(
+            f"lama-cleaner --model {model} --device {device} --hf_access_token={hf_access_token} --port {port}"
+        )
+
diff --git a/scripts/user_scripts/win_config.bat b/scripts/user_scripts/win_config.bat
index 51a72b9..bd9ddb6 100644
--- a/scripts/user_scripts/win_config.bat
+++ b/scripts/user_scripts/win_config.bat
@@ -6,7 +6,8 @@ set PATH=C:\Windows\System32;%PATH%
 
 @call conda-unpack
 
-@call pip3 install -U torch==1.12.1 --extra-index-url https://download.pytorch.org/whl/cu116
+@call conda install -y cudatoolkit=11.3
+@call pip3 install torch --extra-index-url https://download.pytorch.org/whl/cu113
 @call pip3 install -U lama-cleaner
 
 @call invoke config