fix insensible&chrome tasks

evaluation_examples/settings/proxy/dataimpulse.json

@@ -2,8 +2,8 @@
     {
         "host": "gw.dataimpulse.com",
         "port": 823,
-        "username": "e750e5abb74376d28361",
-        "password": "e5ec245537e1e76a",
+        "username": "your_username",
+        "password": "your_password",
         "protocol": "http",
         "provider": "dataimpulse",
         "type": "residential",

mm_agents/openai_cua_agent.py

@@ -753,7 +753,6 @@ class OpenAICUAAgent:
                 # Convert the action to an Action object
                 step_action = Action(action.get("action", ""), self.action_space)
                 # Execute the action in the environment
-                print(f"Executing action: {step_action.get_action()}")
                 obs, reward, terminated, info = self.env.step(step_action.get_action())
                 
                 screenshot_base64 = encode_image(obs["screenshot"])

monitor/.env

@@ -8,7 +8,7 @@ RESULTS_BASE_PATH=../results
 ACTION_SPACE=pyautogui
 OBSERVATION_TYPE=screenshot
 MODEL_NAME=computer-use-preview
-MAX_STEPS=100
+MAX_STEPS=150
 FLASK_PORT=80
 FLASK_HOST=0.0.0.0
 FLASK_DEBUG=false

run_operator.sh

@@ -1,9 +0,0 @@
-python run_multienv_openaicua.py \
---headless \
---observation_type screenshot \
---model computer-use-preview \
---result_dir ./results_operator_full_test_0713 \
---test_all_meta_path evaluation_examples/test_all.json \
---max_steps 100 \
---num_envs 15 \
---provider_name aws

show_result.py

@@ -68,4 +68,4 @@ def get_result(action_space, use_model, observation_type, result_dir):
 
 
 if __name__ == '__main__':
-    get_result("pyautogui", "computer-use-preview", "screenshot", "./results_operator_full_test_0713")
+    get_result("pyautogui", "gpt-4o", "a11y_tree", "./results")