Can’t download "simple-knn" submodule

[YujiaLin-523]

The link to simple-knn in this repo is not working, so that I can't download the submodule.

[YujiaLin-523]

And there is another problem that when I performed evaluation, I met this:

Traceback (most recent call last):
  File "scripts/evaluate.py", line 19, in <module>
    from hugs.trainer import GaussianTrainer
  File "/home/ubuntu/lyj/Project/HUGS/./hugs/trainer/__init__.py", line 6, in <module>
    from .gs_trainer import GaussianTrainer
  File "/home/ubuntu/lyj/Project/HUGS/./hugs/trainer/gs_trainer.py", line 27, in <module>
    from hugs.models.hugs_trimlp import HUGS_TRIMLP
  File "/home/ubuntu/lyj/Project/HUGS/./hugs/models/__init__.py", line 1, in <module>
    from .scene import SceneGS
  File "/home/ubuntu/lyj/Project/HUGS/./hugs/models/scene.py", line 20, in <module>
    from simple_knn._C import distCUDA2
ModuleNotFoundError: No module named 'simple_knn'

I have installed this submodule and installed simple_knn in my conda environment using pip

[LarkLeeOnePiece]

And there is another problem that when I performed evaluation, I met this:

Traceback (most recent call last):
  File "scripts/evaluate.py", line 19, in <module>
    from hugs.trainer import GaussianTrainer
  File "/home/ubuntu/lyj/Project/HUGS/./hugs/trainer/__init__.py", line 6, in <module>
    from .gs_trainer import GaussianTrainer
  File "/home/ubuntu/lyj/Project/HUGS/./hugs/trainer/gs_trainer.py", line 27, in <module>
    from hugs.models.hugs_trimlp import HUGS_TRIMLP
  File "/home/ubuntu/lyj/Project/HUGS/./hugs/models/__init__.py", line 1, in <module>
    from .scene import SceneGS
  File "/home/ubuntu/lyj/Project/HUGS/./hugs/models/scene.py", line 20, in <module>
    from simple_knn._C import distCUDA2
ModuleNotFoundError: No module named 'simple_knn'

I have installed this submodule and installed simple_knn in my conda environment using pip

Could you see the package name"simple_knn" when you used "conda list"?

[YujiaLin-523]

Dear Lark Lee, Yes, I can see it in my conda environment but it still says this module is not installed...[image] Yujia Lin Glasgow College, University of Electronic Science and Technology of China Chengdu, Sichuan

From: "Lark ***@***.***> Date: Fri, Jul 12, 2024, 13:04 Subject: Re: [apple/ml-hugs] Can’t download "simple-knn" submodule (Issue #11) To: ***@***.***> Cc: "Yujia ***@***.***>, "State ***@***.***> > And there is another problem that when I performed evaluation, I met this:Traceback (most recent call last):

File "scripts/evaluate.py", line 19, in <module> from hugs.trainer import GaussianTrainer File "/home/ubuntu/lyj/Project/HUGS/./hugs/trainer/__init__.py", line 6, in <module> from .gs_trainer import GaussianTrainer File "/home/ubuntu/lyj/Project/HUGS/./hugs/trainer/gs_trainer.py", line 27, in <module> from hugs.models.hugs_trimlp import HUGS_TRIMLP File "/home/ubuntu/lyj/Project/HUGS/./hugs/models/__init__.py", line 1, in <module> from .scene import SceneGS File "/home/ubuntu/lyj/Project/HUGS/./hugs/models/scene.py", line 20, in <module> from simple_knn._C import distCUDA2 ModuleNotFoundError: No module named 'simple_knn' I have installed this submodule and installed simple_knn in my conda environment using pip

…

Could you see the package name"simple_knn" when you used "conda list"? — Reply to this email directly, view it on GitHub<#11 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/BB6EBQUZZFOXGRPLS6TJDY3ZL5PV7AVCNFSM6AAAAABKUMRSTOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDEMRUGY2DKMBRGU>. You are receiving this because you modified the open/close state.[image: https://github.com/notifications/beacon/BB6EBQTGWCSQDXOX6VJ34Y3ZL5PV7A5CNFSM6AAAAABKUMRSTOWGG33NNVSW45C7OR4XAZNMJFZXG5LFINXW23LFNZ2KUY3PNVWWK3TUL5UWJTUETFRZO.gif]Message ID: ***@***.***>

[Daydreamer-f]

try this: pip install git+https://gitlab.inria.fr/bkerbl/simple-knn.git

[YujiaLin-523]

try this: pip install git+https://gitlab.inria.fr/bkerbl/simple-knn.git

@Daydreamer-f Well, thank you. But it still not work...

[ZhiyaoZhou]

try this: pip install git+https://gitlab.inria.fr/bkerbl/simple-knn.git

@Daydreamer-f Well, thank you. But it still not work...

or you could try typing "git clone https://gitlab.inria.fr/bkerbl/simple-knn.git" in this directory: ml-hugs/submodules/
then "cd simple-knn" and "pip install ."

[YujiaLin-523]

try this: pip install git+https://gitlab.inria.fr/bkerbl/simple-knn.git

@Daydreamer-f Well, thank you. But it still not work...

or you could try typing "git clone https://gitlab.inria.fr/bkerbl/simple-knn.git" in this directory: ml-hugs/submodules/ then "cd simple-knn" and "pip install ."

I tried this and it still raised the error that there is no simple-knn in my environment. What is your OS and CUDA version? Maybe I can retry it using your settings.
Btw, it should be pip install ./ in your advice rather than pip install.

[Daydreamer-f]

try this: pip install git+https://gitlab.inria.fr/bkerbl/simple-knn.git

@Daydreamer-f Well, thank you. But it still not work...

or you could try typing "git clone https://gitlab.inria.fr/bkerbl/simple-knn.git" in this directory: ml-hugs/submodules/ then "cd simple-knn" and "pip install ."

I tried this and it still raised the error that there is no simple-knn in my environment. What is your OS and CUDA version? Maybe I can retry it using your settings. Btw, it should be pip install ./ in your advice rather than pip install.

I simply type this command in the conda environment which is the same as the one hugs required. I changed torch version to 1.21.0 and CUDA version remain 11.7 to fix the pytorch3d version.
Did you meet any errors when installing simple-knn using pip? And when you installed it successfully, can you try these command in your conda environment:

python
import simple-knn

And report what happened?
(我之前安装simple-knn的时候也遇到过一些报错，比如torch和CUDA版本不符，但都是在安装的时候报的error，simple-knn一般成功安装了运行的时候都没有问题。所以你确定你成功安装了吗？）

[yu-jingrui]

I have a similar problem. Submodules are not installed while running conda_setup.sh. The error message is

RuntimeError:
The detected CUDA version (12.5) mismatches the version that was used to compile
PyTorch (11.7). Please make sure to use the same CUDA versions.

CUDA 12.5 is my system cuda version. I installed cuda 11.7 in the hugs env, but pip does not recognize it and still show the same error.

[yu-jingrui]

I have a similar problem. Submodules are not installed while running conda_setup.sh. The error message is

RuntimeError:
The detected CUDA version (12.5) mismatches the version that was used to compile
PyTorch (11.7). Please make sure to use the same CUDA versions.

CUDA 12.5 is my system cuda version. I installed cuda 11.7 in the hugs env, but pip does not recognize it and still show the same error.

I solved this after consulting claude.ai. CUDA_HOME needs to be specified manually as /home/myusername/miniconda3/envs/hugs. After this, pip reports g++ version is too new (14.x on archlinux) for pytorch-cuda 11.7, which needs to be >6.0 <11.5. Then I conda install gxx_linux-64, which installed gxx_linux-64 version 11.2. This solved the problem. Then I could install the submodules using the commands pip install submodules/simple-knn and pip install submodules/diff-gaussian-rasterization.