The following guidance works well for a machine with 4090 GPU | cuda 11.7 | ubuntu 22.04, a machine with 3090 GPU | cuda 11.6 | ubuntu 20.04, a machine with 3090 GPU | cuda 11.7 | ubuntu 18.04, a machine with 4060 GPU | cuda 11.7 | wsl2, and more machines.
For possible errors, please see ERROR_CATCH.md. Our repo is mainly built upon GNFactor, so you can also refer to GNFactor's installation instruction. If you encounter any other problem, feel free to open an issue.
# for SSH
git clone [email protected]:GuanxingLu/oral.git
# [Optional] We have wrapped (modified) third party packages into this repo, so it might be oversized. To address this, run:
# git config --global http.postBuffer 104857600
conda remove -n manigaussian --all
conda create -n manigaussian python=3.9
conda activate manigaussian
Install pytorch
conda install pytorch==1.10.0 torchvision torchaudio cudatoolkit=11.3 -c pytorch
cd ..
git clone https://github.com/facebookresearch/pytorch3d.git
cd pytorch3d
conda install -c fvcore -c iopath -c conda-forge fvcore iopath
pip install -e .
cd ../ManiGaussian
cd ..
git clone https://github.com/openai/CLIP.git
cd CLIP
pip install -e .
cd ../ManiGaussian
pip install open-clip-torch
Download CoppeliaSim from https://www.coppeliarobotics.com/previousVersions, e.g., CoppeliaSim_Player_V4_1_0_Ubuntu18_04.tar.xz
tar -xvf CoppeliaSim_Player_V4_1_0_Ubuntu18_04.tar.xz
rm CoppeliaSim_Player_V4_1_0_Ubuntu18_04.tar.xz
Remember to source your bashrc (source ~/.bashrc) and reopen a new terminal then.
You should replace the path here with your own path to the coppeliasim installation directory.
export COPPELIASIM_ROOT=EDIT/ME/PATH/TO/COPPELIASIM/INSTALL/DIR
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$COPPELIASIM_ROOT
export QT_QPA_PLATFORM_PLUGIN_PATH=$COPPELIASIM_ROOT
cd third_party/PyRep
pip install -r requirements.txt
pip install .
cd ../..
cd third_party/RLBench
pip install -r requirements.txt
python setup.py develop
cd ../..
cd third_party/YARR
pip install -r requirements.txt
python setup.py develop
cd ../..
pip install -r requirements.txt
pip install packaging==21.3 dotmap pyhocon wandb==0.14.0 chardet opencv-python-headless gpustat ipdb visdom sentencepiece
Install xformers (this version is a must to avoid errors from detectron2)
pip install xformers==0.0.18
Install detectron2:
cd ..
git clone https://github.com/facebookresearch/detectron2.git
cd detectron2
pip install -e .
cd ../ManiGaussian
Install ODISE packages
cd third_party/ODISE
pip install -e .
cd ../..
Since a lot of packages are installed, there are some possible bugs. Use these commands first before running the code.
# update torch
conda install pytorch==2.0.0 torchvision==0.15.0 torchaudio==2.0.0 pytorch-cuda=11.7 -c pytorch -c nvidia
pip install hydra-core==1.1
pip install opencv-python-headless
pip install numpy==1.23.5
cd third_party/gaussian-splatting/
pip install -e submodules/diff-gaussian-rasterization
pip install -e submodules/simple-knn
cd ../..
In ManiGaussian, we use Lightning Fabric to conduct DDP training for Gaussian renderer, rather than the vanilla pytorch DDP. Reference: gaussian-splatting's issue
pip install lightning