Fix some errors found in thunderbird examples

experiment_screenshot_a11y_tree.py

@@ -1,10 +1,10 @@
+import ctypes
 import datetime
 import json
 import logging
 import os
 import sys
-import threading
-import time
+import func_timeout
 
 from desktop_env.envs.desktop_env import DesktopEnv
 from mm_agents.gpt_4v_agent import GPT4v_Agent
@@ -44,9 +44,10 @@ logger.addHandler(sdebug_handler)
 logger = logging.getLogger("desktopenv.experiment")
 
 PATH_TO_VM = r"C:\Users\tianbaox\Documents\Virtual Machines\Ubuntu\Ubuntu.vmx"
-#PATH_TO_VM = "../../../../大文件/镜像/Ubuntu-1218/Ubuntu/Ubuntu.vmx"
 
 
+# PATH_TO_VM = "../../../../大文件/镜像/Ubuntu-1218/Ubuntu/Ubuntu.vmx"
+
 def run_one_example(example, agent, max_steps=10, example_trajectory_dir="exp_trajectory", recording=True):
     trajectory_recording_path = os.path.join(example_trajectory_dir, "trajectory.json")
     env = DesktopEnv(
@@ -105,28 +106,10 @@ def run_one_example(example, agent, max_steps=10, example_trajectory_dir="exp_tr
         except Exception as e:
             print(f"An error occurred while stopping the recording: {e}")
 
-    # Run the `record` function in a separate thread
-    recording_thread = threading.Thread(target=stop_recording())
-    recording_thread.start()
-
-    # Start a timer for your timeout length (in this case, 60 seconds)
-    timeout = 60  # seconds
-    start_time = time.time()
-
-    # The main thread will wait for the set timeout period or until the recording is done
-    while recording_thread.is_alive():
-        elapsed_time = time.time() - start_time
-        if elapsed_time >= timeout:
-            print("Timeout reached. Stopping recording.")
-            break
-        time.sleep(0.1)  # Sleep for a short time to prevent this loop from using too much CPU
-
-    # kill the recording thread if it is still alive
-    if recording_thread.is_alive():
-        recording_thread.kill()
-
-    # Wait for the recording thread to finish before exiting
-    recording_thread.join()
+    try:
+        func_timeout.func_timeout(30, stop_recording)
+    except func_timeout.exceptions.FunctionTimedOut:
+        logger.info("Recording timed out.")
 
     result = env.evaluate()
     logger.info("Result: %.2f", result)
@@ -143,9 +126,9 @@ def run_one_example(example, agent, max_steps=10, example_trajectory_dir="exp_tr
 
 def main(example_class, example_id):
     action_space = "pyautogui"
-    #example_class = "libreoffice_calc"
-    #example_id = "7b6c7e24-c58a-49fc-a5bb-d57b80e5b4c3"
-    #example_id = "01b269ae-2111-4a07-81fd-3fcd711993b0"
+    # example_class = "libreoffice_calc"
+    # example_id = "7b6c7e24-c58a-49fc-a5bb-d57b80e5b4c3"
+    # example_id = "01b269ae-2111-4a07-81fd-3fcd711993b0"
     gpt4_model = "gpt-4-vision-preview"
     gemini_model = "gemini-pro-vision"
 
@@ -155,9 +138,9 @@ def main(example_class, example_id):
 
     with open(f"evaluation_examples/examples/{example_class}/{example_id}.json", "r", encoding="utf-8") as f:
         example = json.load(f)
-    example["snapshot"] = "exp_chrome"
-    #example["snapshot"] = "exp_setup4"
-    #example["snapshot"] = "Snapshot 30"
+    example["snapshot"] = "exp_v1"
+    # example["snapshot"] = "exp_setup4"
+    # example["snapshot"] = "Snapshot 30"
 
     api_key = os.environ.get("OPENAI_API_KEY")
     agent = GPT4v_Agent(api_key=api_key, model=gpt4_model, instruction=example['instruction'],