TypeError: rasterization() got an unexpected keyword argument 'ubp'

[lengfeng-bot]

Thanks for your great job!!
When I want to run the code,an has occurred：

(most recent call last):
  File "spotless_trainer.py", line 1266, in <module>
    main(cfg)
  File "spotless_trainer.py", line 1256, in main
    runner.train()
  File "spotless_trainer.py", line 577, in train
    renders, alphas, info = self.rasterize_splats(
  File "spotless_trainer.py", line 444, in rasterize_splats
    render_colors, render_alphas, info = rasterization(
TypeError: rasterization() got an unexpected keyword argument 'ubp'

When I corrected the gsplat version to match what you said, the same problem still occurred.Finally, I decided to comment out ubp, and the code worked fine.

        render_colors, render_alphas, info = rasterization(
            means=means,
            quats=quats,
            scales=scales,
            opacities=opacities,
            colors=colors,
            viewmats=torch.linalg.inv(camtoworlds),  # [C, 4, 4]
            Ks=Ks,  # [C, 3, 3]
            width=width,
            height=height,
            packed=self.cfg.packed,
            absgrad=self.cfg.absgrad,
            #ubp=self.cfg.ubp,
            sparse_grad=self.cfg.sparse_grad,
            rasterize_mode=rasterize_mode,
            **kwargs,
        )
        return render_colors, render_alphas, info

But when I added the ubp parameter, the code gave an error, as shown below：

Traceback (most recent call last):
  File "spotless_trainer.py", line 1269, in <module>
    main(cfg)
  File "spotless_trainer.py", line 1259, in main
    runner.train()
  File "spotless_trainer.py", line 727, in train
    self.update_running_stats(info)
  File "/home/deepnavi/abandon/conda_envs/gaussian/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context
    return func(*args, **kwargs)
  File "spotless_trainer.py", line 887, in update_running_stats
    sqrgrads = info["means2d"].sqrgrad.clone()
AttributeError: 'Tensor' object has no attribute 'sqrgrad'
^CError in atexit._run_exitfuncs:
Traceback (most recent call last):
  File "/home/deepnavi/abandon/conda_envs/gaussian/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1504, in _clean_up_worker
    w.join(timeout=_utils.MP_STATUS_CHECK_INTERVAL)
  File "/home/deepnavi/abandon/conda_envs/gaussian/lib/python3.8/multiprocessing/process.py", line 149, in join
    res = self._popen.wait(timeout)
  File "/home/deepnavi/abandon/conda_envs/gaussian/lib/python3.8/multiprocessing/popen_fork.py", line 44, in wait
    if not wait([self.sentinel], timeout):
  File "/home/deepnavi/abandon/conda_envs/gaussian/lib/python3.8/multiprocessing/connection.py", line 931, in wait
    ready = selector.select(timeout)
  File "/home/deepnavi/abandon/conda_envs/gaussian/lib/python3.8/selectors.py", line 415, in select
    fd_event_list = self._selector.poll(timeout)
``
I would like to know how to solve this problem

[HanLingsgjk]

Firstly, you need to confirm whether your gsplat version is the same as this project, which seems to be using version 1.1.1.

Secondly, just "pip install gsplat" is not enough. It also needs to be installed from the source code. I think the author should provide more detailed information

[lilygoli]

Hi!
It is probably due not having the changed CUDA kernel installed. The CUDA kernel in this repository is changed minimally from the original splat to include the UBP feature. If you have the splat library installed already, you can reinstall the kernel through running the following command in the main directory:

python setup.py clean --all
python setup.py install

Otherwise, please install the library from the source:

pip install git+https://github.com/lilygoli/SpotLessSplats.git

[lengfeng-bot]

First of all, thank you for your suggestion. I think this is a good solution. I didn't know much about the gsplat framework before. It is a very convenient integrated framework. After modifying the CUDA function, it seems that we should to use the setup.py command to install our own version.
When I run:

python setup.py clean --all
python setup.py install

The program can run normally.
Thank you again for your great work!