in subprocess can't use cuda

[nuclear-missile]

Checklist

• I have searched for similar issues.
• For Python issues, I have tested with the latest development wheel.
• I have checked the release documentation and the latest documentation (for main branch).

Describe the issue

when I import open3d, subprocess can't use cuda
while without import open3d, it runs

Steps to reproduce the bug

import torch
from torch.multiprocessing import Process
import open3d # without this, cuda both available in main process and subprocess

class MyProcess(Process):
    def __init__(self):
        super(MyProcess, self).__init__()

    def run(self):
        print('subprocess:', torch.cuda.is_available())


if __name__ == '__main__':
    p = MyProcess()
    p.start()
    print('mainprocess:', torch.cuda.is_available())

Error message

mainprocess: True
/root/miniconda3/envs/dp3/lib/python3.8/site-packages/torch/cuda/init.py:128: UserWarning: CUDA initialization: CUDA driver initialization failed, you might not have a CUDA gpu. (Triggered internally at /opt/conda/conda-bld/pytorch_1724789115405/work/c10/cuda/CUDAFunctions.cpp:108.)
return torch._C._cuda_getDeviceCount() > 0
subprocess: False

Expected behavior

mainprocess: True
subprocess: True

Open3D, Python and System information

- Operating system: Ubuntu 20.04
- Python version: Python 3.8.20
- Open3D version: output from python: 0.18.0
- System architecture: x86
- Is this a remote workstation?: yes
- How did you install Open3D?: both tried on conda & build from source
- Compiler version (if built from source): gcc 9.4.0

Additional information

pytorch: 2.4.1
NVIDIA-SMI 550.107.02
Driver Version: 550.107.02
CUDA Version: 12.4