AttributeError: 'ConfigDict' object has no attribute 'train_dir' in SemanticSegmentation

### Checklist

- [X] I have searched for [similar issues](https://github.com/isl-org/Open3D-ML/issues).
- [X] I have tested with the [latest development wheel](http://www.open3d.org/docs/latest/getting_started.html#development-version-pip).
- [X] I have checked the [release documentation](http://www.open3d.org/docs/release/) and the [latest documentation](http://www.open3d.org/docs/latest/) (for `main` branch).

### Describe the issue

I am trying to perform semantic segmentation but a lot of errors occur while doing that.
Specifically, something is wrong with the definition`open3d.ml.torch.datasets.Custom3D` because when I put it as parameter in `SemanticSegmentation` I am getting `AttributeError: 'ConfigDict' object has no attribute 'train_dir'`, which resides in `customdataset.py` into `Custom3D` class. My input in `open3d.ml.torch.datasets.Custom3D` is the pcd path.
Do you anything about that?

### Steps to reproduce the bug

```python
import sys

from zmq import device
sys.path.append('/home/johnny/Open3D-ML')
from ml3d.torch.dataloaders import get_sampler, TorchDataloader, DefaultBatcher, ConcatBatcher
from ml3d.torch.pipelines import SemanticSegmentation
from open3d.ml.torch.models import RandLANet
import open3d.ml as o3dml

import numpy as np
import open3d as o3d

pcd_path = 'path/to/pcd'
custom_dataset = o3dml.torch.datasets.Custom3D(dataset_path=pcd_path, name='real_complete_pcd')
model = RandLANet(dim_input=3)
custom_model = SemanticSegmentation(
model= model,
dataset=custom_dataset,
device='gpu'
)
```

### Error message

_No response_

### Expected behavior

_No response_

### Open3D, Python and System information

```markdown
- Operating system: Ubuntu 22.04
- Python version: python 3.10.12
- Open3D version: 0.18.0
- System type: x64
- Is this remote workstation?: no
- How did you install Open3D?: pip
- Compiler version: None
```

### Additional information

_No response_