Skip to content

Commit 41c4ead

Browse files
Merge pull request #263 from ServiceNow/osworld
Update relaunch flag to False and adjust model_args for OSWORLD_OAI
2 parents 684f56b + 89c2ca8 commit 41c4ead

File tree

2 files changed

+4
-7
lines changed

2 files changed

+4
-7
lines changed

experiments/run_osworld.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -28,7 +28,7 @@ def get_task_ids() -> set[str]:
2828
def main():
2929
n_jobs = 4
3030
use_vmware = True
31-
relaunch = True
31+
relaunch = False
3232
agent_args = [
3333
OSWORLD_CLAUDE,
3434
# OSWORLD_OAI # performs poorly.

src/agentlab/agents/tool_use_agent/tool_use_agent.py

Lines changed: 3 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -7,21 +7,18 @@
77
from typing import Any
88

99
import bgym
10-
import numpy as np
1110
import pandas as pd
11+
from bgym import Benchmark as BgymBenchmark
1212
from browsergym.core.observation import extract_screenshot
1313
from browsergym.utils.obs import (
1414
flatten_axtree_to_str,
1515
flatten_dom_to_str,
1616
overlay_som,
1717
prune_html,
1818
)
19-
from PIL import Image
2019

21-
from agentlab.agents import agent_utils
22-
from agentlab.benchmarks.abstract_env import AbstractBenchmark as AgentLabBenchmark
23-
from bgym import Benchmark as BgymBenchmark
2420
from agentlab.agents.agent_args import AgentArgs
21+
from agentlab.benchmarks.abstract_env import AbstractBenchmark as AgentLabBenchmark
2522
from agentlab.benchmarks.osworld import OSWorldActionSet
2623
from agentlab.llm.base_api import BaseModelArgs
2724
from agentlab.llm.llm_utils import image_to_png_base64_url
@@ -629,7 +626,7 @@ def get_action(self, obs: Any) -> float:
629626
)
630627

631628
OSWORLD_OAI = ToolUseAgentArgs(
632-
model_args=OPENAI_MODEL_CONFIG,
629+
model_args=GPT_4_1_MINI,
633630
config=PromptConfig(
634631
tag_screenshot=True,
635632
goal=Goal(goal_as_system_msg=True),

0 commit comments

Comments
 (0)