lzy/sci-gui-agent-benchmark
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: add flexible recording control and improve execution logging

Browse Source
This commit is contained in:
cui0711
2026-01-30 16:28:13 +08:00
parent 47bcfc0f0b
commit 716d82f4d1
4 changed files with 67 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
lib_run_single.py
View File

@@ -13,30 +13,44 @@ def run_single_example(agent, env, example, max_steps, instruction, args, exampl
runtime_logger = setup_logger(example, example_result_dir)
# Reset environment first to get fresh VM IP
env.reset(task_config=example)
# env.reset(task_config=example)
# logger.info("=======Environment reset completed=======")
# Reset agent with fresh VM IP (for snapshot reverts)
try:
agent.reset(runtime_logger, vm_ip=env.vm_ip)
except Exception as e:
agent.reset(vm_ip=env.vm_ip)
# # Reset agent with fresh VM IP (for snapshot reverts)
# try:
# agent.reset(runtime_logger, vm_ip=env.vm_ip)
# except Exception as e:
# agent.reset(vm_ip=env.vm_ip)
time.sleep(60) # Wait for the environment to be ready
# time.sleep(10) # Wait for the environment to be ready
# get initial observation
logger.info("Getting initial observation...")
obs = env._get_obs() # Get the initial observation
logger.info("Initial observation obtained.")
done = False
step_idx = 0
env.controller.start_recording()
if getattr(args, 'enable_recording', False):
env.controller.start_recording()
while not done and step_idx < max_steps:
logger.info(f"Step {step_idx + 1} prediction...")
response, actions = agent.predict(
instruction,
obs
)
logger.info(f"Response: {response}")
logger.info(f"Actions: {actions}")
logger.info(f"Executing actions...")
for action in actions:
# Capture the timestamp before executing the action
action_timestamp = datetime.datetime.now().strftime("%Y%m%d@%H%M%S%f")
logger.info("Step %d: %s", step_idx + 1, action)
logger.info("执行动作中...")
obs, reward, done, info = env.step(action, args.sleep_after_execution)
logger.info("动作执行完成。")
logger.info("Reward: %.2f", reward)
logger.info("Done: %s", done)
# Save screenshot and trajectory information
@@ -69,7 +83,8 @@ def run_single_example(agent, env, example, max_steps, instruction, args, exampl
# Log task completion to results.json
log_task_completion(example, result, example_result_dir, args)
env.controller.end_recording(os.path.join(example_result_dir, "recording.mp4"))
if getattr(args, 'enable_recording', False):
env.controller.end_recording(os.path.join(example_result_dir, "recording.mp4"))
def setup_logger(example, example_result_dir):