关于osworld测试

[icey-zhang]

我想用qwen-vl-32b模型（已本地部署并通过litellm进行转发）去测试osworld测试套，请问我的测试指令应该怎么写？
这样是正确的么？./run-local.sh --mode dev --json-name "evaluation_examples_windows/test_osworld.json" --agent UFO --agent-settings '{"llm_type": "openai", "llm_endpoint": "https:localhost:8008/v1/chat/completions", "llm_auth": {"type": "api-key", "token": ""}}'

另外这样的指令进行测试时会遇到这个错误
Agent response: {'message': "[Errno 2] No such file or directory: '\\\\host.lan\\Data\\mm_agents\\UFO\\logs\\excel/4188d3a4-077d-46b7-9c86-23e1a036f6c1\\evaluation.log'", 'status': 'error'}

请问应该如何正确设置其他模型的测试指令

[nice-mee]

This is a bit weird, the command should be correct. Have you tested other models like gpt-4o or gemini-2.5-flash?

The error given seems unrelated to LLM deployment. I suspect it's probably because some configurations are not done yet. Can you take a look at windows_arena/Ufo-runningLogs.txt under UFO folder?