99 lines
2.3 KiB
JSON
99 lines
2.3 KiB
JSON
{
|
|
"id": "5c1075ca-bb34-46a3-a7a0-029bd7463e79",
|
|
"snapshot": "os",
|
|
"instruction": "Copy all files matching \"*failed.ipynb\" in the current directory tree to \"./fails\" preserving the directory hierarchy",
|
|
"source": "NL2Bash",
|
|
"config": [
|
|
{
|
|
"type": "download",
|
|
"parameters": {
|
|
"files": [
|
|
{
|
|
"url": "https://huggingface.co/datasets/xlangai/ubuntu_osworld_file_cache/resolve/main/os/5c1075ca-bb34-46a3-a7a0-029bd7463e79/setup.sh",
|
|
"path": "setup.sh"
|
|
}
|
|
]
|
|
}
|
|
},
|
|
{
|
|
"type": "execute",
|
|
"parameters": {
|
|
"command": "chmod +x setup.sh",
|
|
"shell": true
|
|
}
|
|
},
|
|
{
|
|
"type": "download",
|
|
"parameters": {
|
|
"files": [
|
|
{
|
|
"url": "https://huggingface.co/datasets/xlangai/ubuntu_osworld_file_cache/resolve/main/os/5c1075ca-bb34-46a3-a7a0-029bd7463e79/eval.sh",
|
|
"path": "eval.sh"
|
|
}
|
|
]
|
|
}
|
|
},
|
|
{
|
|
"type": "execute",
|
|
"parameters": {
|
|
"command": "chmod +x eval.sh",
|
|
"shell": true
|
|
}
|
|
},
|
|
{
|
|
"type": "execute",
|
|
"parameters": {
|
|
"command": "./setup.sh",
|
|
"shell": true
|
|
}
|
|
},
|
|
{
|
|
"type": "execute",
|
|
"parameters": {
|
|
"command": [
|
|
"python",
|
|
"-c",
|
|
"import pyautogui; import time; pyautogui.click(960, 540); time.sleep(0.5); pyautogui.hotkey('ctrl', 'alt', 't'); time.sleep(0.5)"
|
|
]
|
|
}
|
|
},
|
|
{
|
|
"type": "activate_window",
|
|
"parameters": {
|
|
"window_name": "Terminal"
|
|
}
|
|
},
|
|
{
|
|
"type": "execute",
|
|
"parameters": {
|
|
"command": [
|
|
"python",
|
|
"-c",
|
|
"import pyautogui; import time; time.sleep(0.5); pyautogui.write('cd test_environment'); time.sleep(0.5); pyautogui.press('enter')"
|
|
]
|
|
}
|
|
}
|
|
],
|
|
"trajectory": "trajectories/",
|
|
"related_apps": [
|
|
"os"
|
|
],
|
|
"evaluator": {
|
|
"func": "check_include_exclude",
|
|
"result": {
|
|
"type": "vm_command_line",
|
|
"command": "bash eval.sh",
|
|
"shell": true
|
|
},
|
|
"expected": {
|
|
"type": "rule",
|
|
"rules": {
|
|
"include": [
|
|
"Evaluation successful."
|
|
],
|
|
"exclude": []
|
|
}
|
|
}
|
|
},
|
|
"proxy": false
|
|
} |