ImportError: cannot import name 'csrc' from 'gsplat'

[sblat021]

Whenever I run ns-train with gaussian-splatting or igs2gs I get the following error (it works fine when using nerfacto):

---

{NOTE] Not running eval iterations since only viewer is enabled.
Use --vis {wandb, tensorboard, viewer+wandb, viewer+tensorboard} to run with eval.
No Nerfstudio checkpoint to load, so training from scratch.
Disabled comet/tensorboard/wandb event writers
C:\Users\micro\anaconda3\envs\nerfstudio\lib\site-packages\torch\utils\cpp_extension.py:359: UserWarning: Error checking compiler version for cl:
[WinError 2] The system cannot find the file specified
warnings.warn(f'Error checking compiler version for {compiler}: {error}')
( ● ) gsplat: Setting up CUDA (This may take a few minutes the first time)INFO: Could not find files for the given pattern(s).
Printing profiling stats, from longest to shortest duration in seconds
VanillaPipeline.get_train_loss_dict: 0.6218
Trainer.train_iteration: 0.6218
Traceback (most recent call last):
File "C:\Users\micro\anaconda3\envs\nerfstudio\lib\site-packages\gsplat\cuda_backend.py", line 56, in
from gsplat import csrc as C
ImportError: cannot import name 'csrc' from 'gsplat' (C:\Users\micro\anaconda3\envs\nerfstudio\lib\site-packages\gsplat_init.py)

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "C:\Users\micro\anaconda3\envs\nerfstudio\lib\runpy.py", line 194, in _run_module_as_main
return _run_code(code, main_globals, None,
File "C:\Users\micro\anaconda3\envs\nerfstudio\lib\runpy.py", line 87, in run_code
exec(code, run_globals)
File "C:\Users\micro\anaconda3\envs\nerfstudio\Scripts\ns-train.exe_main.py", line 7, in
File "C:\Apps\nerfstudio\nerfstudio\scripts\train.py", line 262, in entrypoint
main(
File "C:\Apps\nerfstudio\nerfstudio\scripts\train.py", line 247, in main
launch(
File "C:\Apps\nerfstudio\nerfstudio\scripts\train.py", line 189, in launch
main_func(local_rank=0, world_size=world_size, config=config)
File "C:\Apps\nerfstudio\nerfstudio\scripts\train.py", line 100, in train_loop
trainer.train()
File "C:\Apps\nerfstudio\nerfstudio\engine\trainer.py", line 250, in train
loss, loss_dict, metrics_dict = self.train_iteration(step)
File "C:\Apps\nerfstudio\nerfstudio\utils\profiler.py", line 127, in inner
out = func(*args, **kwargs)
File "C:\Apps\nerfstudio\nerfstudio\engine\trainer.py", line 471, in train_iteration
_, loss_dict, metrics_dict = self.pipeline.get_train_loss_dict(step=step)
File "C:\Apps\nerfstudio\nerfstudio\utils\profiler.py", line 127, in inner
out = func(*args, **kwargs)
File "C:\Apps\nerfstudio\nerfstudio\pipelines\base_pipeline.py", line 304, in get_train_loss_dict
model_outputs = self._model(ray_bundle) # train distributed data parallel model if world_size > 1
File "C:\Users\micro\anaconda3\envs\nerfstudio\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "C:\Apps\nerfstudio\nerfstudio\models\base_model.py", line 143, in forward
return self.get_outputs(ray_bundle)
File "C:\Apps\nerfstudio\nerfstudio\models\gaussian_splatting.py", line 591, in get_outputs
self.xys, depths, self.radii, conics, num_tiles_hit, _ = ProjectGaussians.apply( # type: ignore
File "C:\Users\micro\anaconda3\envs\nerfstudio\lib\site-packages\torch\autograd\function.py", line 506, in apply
return super().apply(*args, **kwargs) # type: ignore[misc]
File "C:\Users\micro\anaconda3\envs\nerfstudio\lib\site-packages\gsplat\project_gaussians.py", line 67, in forward
) = C.project_gaussians_forward(
File "C:\Users\micro\anaconda3\envs\nerfstudio\lib\site-packages\gsplat\cuda_init.py", line 7, in call_cuda
from ._backend import _C
File "C:\Users\micro\anaconda3\envs\nerfstudio\lib\site-packages\gsplat\cuda_backend.py", line 79, in
_C = load(
File "C:\Users\micro\anaconda3\envs\nerfstudio\lib\site-packages\torch\utils\cpp_extension.py", line 1284, in load
return _jit_compile(
File "C:\Users\micro\anaconda3\envs\nerfstudio\lib\site-packages\torch\utils\cpp_extension.py", line 1509, in _jit_compile
_write_ninja_file_and_build_library(
File "C:\Users\micro\anaconda3\envs\nerfstudio\lib\site-packages\torch\utils\cpp_extension.py", line 1611, in _write_ninja_file_and_build_library
_write_ninja_file_to_build_library(
File "C:\Users\micro\anaconda3\envs\nerfstudio\lib\site-packages\torch\utils\cpp_extension.py", line 2048, in _write_ninja_file_to_build_library
_write_ninja_file(
File "C:\Users\micro\anaconda3\envs\nerfstudio\lib\site-packages\torch\utils\cpp_extension.py", line 2188, in _write_ninja_file
cl_paths = subprocess.check_output(['where',
File "C:\Users\micro\anaconda3\envs\nerfstudio\lib\subprocess.py", line 415, in check_output
return run(*popenargs, stdout=PIPE, timeout=timeout, check=True,
File "C:\Users\micro\anaconda3\envs\nerfstudio\lib\subprocess.py", line 516, in run
raise CalledProcessError(retcode, process.args,
subprocess.CalledProcessError: Command '['where', 'cl']' returned non-zero exit status 1.

Steps to reproduce the behavior:

1. After running COLMAP I run : ns-train gaussian-splatting --data output
2. First few steps seem to work fine and then it crashes once it gets to: Trainer.train_iteration
3. See error message above

Expected behavior
I expect the COLMAP data to be trained using the gaussian-splatting or igs2gs model.

Additional context
I followed the installation instructions exactly.

After nerfstudio dependencies I installed gsplat using:
pip install git+https://github.com/nerfstudio-project/gsplat.git

After that I installed nerfstudio following instructions:
git clone https://github.com/nerfstudio-project/nerfstudio.git
cd nerfstudio
pip install --upgrade pip setuptools
pip install -e .

Visual Studio 2019 is installed including Desktop development with C+.

[sblat021]

This was resolved by uninstalling gsplat and reinstalling it like so:
pip install git+https://github.com/nerfstudio-project/gsplat.git@v0.1.0

[Lanig77]

hi, what is the command to remove previous gsplat install. I have exactly the same issue

[javieryu]

Worked for me.

pip uninstall gsplat
pip install git+https://github.com/nerfstudio-project/gsplat.git@v0.1.0

[vincenzorin]

If the error persists, consider installing from

pip install git+https://github.com/nerfstudio-project/gsplat.git

as suggested in #2685 (comment)

[brentyi]

If you're seeing this issue in PyTorch 2.0, there's also a good chance it's a PyTorch bug. Upgrading to 2.1 fixes it for some systems.

(#2748 has notes on why)

[ILLCamaras]

I have PyTorch 2.1.2+cu118 installed

and I've tried:
pip uninstall gsplat
pip install git+https://github.com/nerfstudio-project/gsplat.git@v0.1.0.

and I get

...
Installing collected packages: gsplat
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
nerfstudio 1.0.0 requires gsplat==0.1.2.1, but you have gsplat 0.1.2 which is incompatible.

and also

pip uninstall gsplat
pip install git+https://github.com/nerfstudio-project/gsplat.git@v0.1.2.1.

and always

Collecting git+https://github.com/nerfstudio-project/gsplat.git@v0.1.2.1.
Cloning https://github.com/nerfstudio-project/gsplat.git (to revision v0.1.2.1.) to c:\users\carlo\appdata\local\temp\pip-req-build-7rgbikln
Running command git clone --filter=blob:none --quiet https://github.com/nerfstudio-project/gsplat.git 'C:\Users\carlo\AppData\Local\Temp\pip-req-build-7rgbikln'
WARNING: Did not find branch or tag 'v0.1.2.1.', assuming revision or ref.' 'Running command git checkout -q v0.1.2.1. error: pathspec 'v0.1.2.1.' did not match any file(s) known to git error: subprocess-exited-with-error

× git checkout -q v0.1.2.1. did not run successfully.
│ exit code: 1
╰─> See above for output.

note: This error originates from a subprocess, and is likely not a problem with pip. error: subprocess-exited-with-error

× git checkout -q v0.1.2.1. did not run successfully. │ exit code: 1 ╰─> See above for output.

note: This error originates from a subprocess, and is likely not a problem with pip.

[brentyi]

Do you have an extra period here?

pip install git+https://github.com/nerfstudio-project/gsplat.git@v0.1.2.1. => pip install git+https://github.com/nerfstudio-project/gsplat.git@v0.1.2.1 may work.

[ILLCamaras]

Install apparently work with only a Yellow warning:
WARNING: Failed to remove contents in a temporary directory 'C:\Users\carlo\AppData\Local\Temp\pip-req-build-w_szq5df'.
You can safely remove it manually.

But it doesn't and I can't run ns-train either as splatfacto
or gaussian-splatting

Here the splatfacto error if could help

`Traceback (most recent call last):
File "C:\Users\carlo\anaconda3\envs\nerfstudio\lib\runpy.py", line 194, in _run_module_as_main
return _run_code(code, main_globals, None,
File "C:\Users\carlo\anaconda3\envs\nerfstudio\lib\runpy.py", line 87, in run_code
exec(code, run_globals)
File "C:\Users\carlo\anaconda3\envs\nerfstudio\Scripts\ns-train.exe_main.py", line 4, in
File "C:\Users\carlo\nerfstudio\nerfstudio\scripts\train.py", line 62, in
from nerfstudio.configs.method_configs import AnnotatedBaseConfigUnion
File "C:\Users\carlo\nerfstudio\nerfstudio\configs\method_configs.py", line 28, in
from nerfstudio.configs.external_methods import get_external_methods
File "C:\Users\carlo\nerfstudio\nerfstudio\configs\external_methods.py", line 24, in
from nerfstudio.engine.trainer import TrainerConfig
File "C:\Users\carlo\nerfstudio\nerfstudio\engine\trainer.py", line 39, in
from nerfstudio.viewer_legacy.server.viewer_state import ViewerLegacyState
File "C:\Users\carlo\nerfstudio\nerfstudio\viewer_legacy\server\viewer_state.py", line 41, in
from nerfstudio.viewer_legacy.server.render_state_machine import RenderAction, RenderStateMachine
File "C:\Users\carlo\nerfstudio\nerfstudio\viewer_legacy\server\render_state_machine.py", line 27, in
from nerfstudio.models.gaussian_splatting import GaussianSplattingModel
File "C:\Users\carlo\nerfstudio\nerfstudio\models\gaussian_splatting.py", line 30, in
from gsplat.compute_cumulative_intersects import compute_cumulative_intersects
ModuleNotFoundError: No module named 'gsplat.compute_cumulative_intersects'

[brentyi]

Are you on the latest version of nerfstudio?

[ILLCamaras]

That's my intention, I've completely rebuilt the anaconda environment installing:

git clone https://github.com/nerfstudio-project/nerfstudio.git
cd nerfstudio
pip install --upgrade pip setuptools
pip install -e .

[brentyi]

Interesting, what's strange to me is this part of your error:

File "C:\Users\carlo\nerfstudio\nerfstudio\models\gaussian_splatting.py", line 30, in
from gsplat.compute_cumulative_intersects import compute_cumulative_intersects
ModuleNotFoundError: No module named 'gsplat.compute_cumulative_intersects'

We no longer have a gaussian_splatting.py in the repo, nor is the line that's producing this error present in the main branch when I search for it. Are you still getting this error despite a fresh clone of the repo?

[ILLCamaras]

I was rebuilding everything again but while installing it asked me to upgrade anaconda, I did it and I've lost ability to run Anaconda Navigator and installation went wrong. I'm uninstalling everything and start again from scratch. Thanks for help, I let you know.

[ILLCamaras]

I've reinstalled everything from Anaconda down to Nerfstudio following the guide but
`ns-train splatfacto --data C:\Users\carlo\3DGSprojects\1750_CM --pipeline.model.cull-alpha-thresh 0.002 --max-num-iterations 100000 colmap --load_3D_points=True --downscale_factor 2'
give me back the same error as before at first run, so I came back here and run

pip uninstall gsplat
pip install git+https://github.com/nerfstudio-project/gsplat.git@v0.1.0

and also

pip uninstall gsplat
pip install git+https://github.com/nerfstudio-project/gsplat.git

receiving the same error.

So I close the shell and run again the same command in a new one and now it's running!
I'm around 15% now, it will take another 3h to complete.
I'll post the results.

[ichsan2895]

Now, please run

conda env export --no-builds > environment-nerfstudio.yml

so, in the next time, if (someday) your system breaks and you must reinstall python/conda, you can replicate the previous environment by running:

conda env create -f environment-nerfstudio.yml -n nerfstudio