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

fix(runner): pass result_dir to evaluate and re-enable environment reset

Browse Source
This commit is contained in:
cui0711
2026-02-05 16:55:49 +08:00
parent ad46acc5f3
commit 63484c7b7b
1 changed files with 8 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
lib_run_single.py
View File

@@ -13,8 +13,8 @@ def run_single_example(agent, env, example, max_steps, instruction, args, exampl
runtime_logger = setup_logger(example, example_result_dir) runtime_logger = setup_logger(example, example_result_dir)
# Reset environment first to get fresh VM IP # Reset environment first to get fresh VM IP
# env.reset(task_config=example) env.reset(task_config=example)
# logger.info("=======Environment reset completed=======") logger.info("=======Environment reset completed=======")
# # Reset agent with fresh VM IP (for snapshot reverts) # # Reset agent with fresh VM IP (for snapshot reverts)
# try: # try:
@@ -74,8 +74,7 @@ def run_single_example(agent, env, example, max_steps, instruction, args, exampl
break break
step_idx += 1 step_idx += 1
time.sleep(20) # Wait for the environment to settle time.sleep(20) # Wait for the environment to settle
result = env.evaluate() result = env.evaluate(result_dir=example_result_dir)
logger.info("Result: %.2f", result)
scores.append(result) scores.append(result)
with open(os.path.join(example_result_dir, "result.txt"), "w", encoding="utf-8") as f: with open(os.path.join(example_result_dir, "result.txt"), "w", encoding="utf-8") as f:
f.write(f"{result}\n") f.write(f"{result}\n")
@@ -112,7 +111,7 @@ def run_single_example_human(env, example, max_steps, instruction, args, example
f.write("\n") f.write("\n")
# Evaluate the result # Evaluate the result
result = env.evaluate() result = env.evaluate(result_dir=example_result_dir)
logger.info("Result: %.2f", result) logger.info("Result: %.2f", result)
scores.append(result) scores.append(result)
with open(os.path.join(example_result_dir, "result.txt"), "w", encoding="utf-8") as f: with open(os.path.join(example_result_dir, "result.txt"), "w", encoding="utf-8") as f:
@@ -549,7 +548,7 @@ def run_single_example_os_symphony(agent, env, example, max_steps, instruction,
break break
step_idx += 1 step_idx += 1
end_time = time.time() end_time = time.time()
result = float(env.evaluate()) result = float(env.evaluate(result_dir=example_result_dir))
logger.info("Result: %.2f", result) logger.info("Result: %.2f", result)
scores.append(result) scores.append(result)
with open(os.path.join(example_result_dir, "result.txt"), "w", encoding="utf-8") as f: with open(os.path.join(example_result_dir, "result.txt"), "w", encoding="utf-8") as f:
@@ -647,7 +646,7 @@ def run_single_example_evocua(agent, env, example, max_steps, instruction, args,
step_idx += 1 step_idx += 1
time.sleep(20) # Wait for environment to settle time.sleep(20) # Wait for environment to settle
result = env.evaluate() result = env.evaluate(result_dir=example_result_dir)
logger.info("Result: %.2f", result) logger.info("Result: %.2f", result)
scores.append(result) scores.append(result)