neus-facto training fails on SDF studio data: _LazyError object has no attribute __data

[kiranganeshan]

When running ns-train neus-facto with the sdfstudio-data dataloader, just after training and evaluation images are cached, an AttributeError occurs: AttributeError: '_LazyError' object has no attribute '__data'.

To Reproduce
Steps to reproduce the behavior:

1. Generate a NeRF Studio dataset from a previous COLMAP run using ns-process-data images --skip-colmap --colmap-model-path <MODEL_PATH> --data <DATA_PATH> --output_dir <NERFSTUDIO_PATH>
2. Clone the sdfstudio repository at https://github.com/autonomousvision/sdfstudio
3. Generate an SDFStudio dataset using sdfstudio/scripts/datasets/process_nerfstudio_to_sdfstudio.py --data <NERFSTUDIO_PATH> --output_dir <SDFSTUDIO_PATH>
4. Run ns-train neus-facto sdfstudio-data <SDFSTUDIO_PATH>

Expected behavior
Output is as follows after config is printed:

/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/engine/trainer.py:137: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead.
  self.grad_scaler = GradScaler(enabled=self.use_grad_scaler)
           Saving checkpoints to: outputs/unnamed/neus-facto/2026-04-02_205403/nerfstudio_models          trainer.py:142
Setting up training dataset...
Caching all 165 images.
Setting up evaluation dataset...
Caching all 165 images.
Traceback (most recent call last):
  File "/Portal/anaconda3/envs/sdfstudio/bin/ns-train", line 6, in <module>
    sys.exit(entrypoint())
  File "/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/scripts/train.py", line 262, in entrypoint
    main(
  File "/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/scripts/train.py", line 247, in main
    launch(
  File "/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/scripts/train.py", line 189, in launch
    main_func(local_rank=0, world_size=world_size, config=config)
  File "/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/scripts/train.py", line 99, in train_loop
    trainer.setup()
  File "/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/engine/trainer.py", line 158, in setup
    self.pipeline = self.config.pipeline.setup(
  File "/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/configs/base_config.py", line 53, in setup
    return self._target(self, **kwargs)
  File "/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/pipelines/base_pipeline.py", line 270, in __init__
    self._model = config.model.setup(
  File "/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/configs/base_config.py", line 53, in setup
    return self._target(self, **kwargs)
  File "/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/models/base_model.py", line 85, in __init__
    self.populate_modules()  # populate the modules
  File "/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/models/neus_facto.py", line 89, in populate_modules
    super().populate_modules()
  File "/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/models/neus.py", line 59, in populate_modules
    super().populate_modules()
  File "/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/models/base_surface_model.py", line 98, in populate_modules
    self.field = self.config.sdf_field.setup(
  File "/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/configs/base_config.py", line 53, in setup
    return self._target(self, **kwargs)
  File "/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/fields/sdf_field.py", line 150, in __init__
    self.encoding = tcnn.Encoding(
  File "/Portal/anaconda3/envs/sdfstudio/lib/python3.10/site-packages/nerfstudio/utils/external.py", line 35, in __getattr__
    return _LazyError.LazyErrorObj(object.__getattribute__(self, "__data"))
AttributeError: '_LazyError' object has no attribute '__data'

Additional context
To my understanding, I cannot use ns-train neus-facto nerfstudio-data <NERFSTUDIO_DATA> because this pipeline requires normals. Meanwhile, ns-train neus-facto sdfstudio-data does not seem to require normals. I'm not sure whether autonomousvision/sdfstudio is the right repository to translate nerfstudio data into sdfstudio data.