search

graphdeco-inria / gaussian-splatting

Original reference implementation of "3D Gaussian Splatting for Real-Time Radiance Field Rendering"
https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/
Other
13.22k stars 1.67k forks source link

A possible reason why installing submodules diff-gaussian-rasterization and simple-knn failed #727

Open Jingfan-Wang opened 4 months ago

Jingfan-Wang commented 4 months ago

FU CK the conda install command！！！

If you have a similar situation：

  1. No matter what version of python, torch, or cuda is installed, the submodule cannot be successfully installed.
  2. The error prompt for installing submodules is related to the CUDA version or CUDA environment. Similar to this： No CUDA runtime is found, using CUDA_HOME='/home/xxx/anaconda3/pkgs/cuda-nvcc-11.8.89-0'

This may be because you used the conda install cudatoolkit or created the environment using a. yml file when using the Conda virtual environment. The installation path for conda install is similar to “/home/xxx/anaconda3/pkgs/cuda-nvcc-11.8.89-0”，instead of/usr/local/cuda-xx.x/bin/nvcc.

Even using a soft link to make the nvcc path point to the actual installation path doesn't help. You need to install cuda manually （cuda12.1）： wget https://developer.download.nvidia.com/compute/cuda/12.1.1/local_installers/cuda_12.1.1_530.30.02_linux.run sudo sh cuda_12.1.1_530.30.02_linux.run

Aseeef commented 4 months ago

If you just git cloned this repo, then submodules/simple-knn and diff-gaussian-rasterization are not included by default. You need to clone them into the correct folders too.