Skip to content

Latest commit

 

History

History
60 lines (50 loc) · 2.68 KB

INSTALL.md

File metadata and controls

60 lines (50 loc) · 2.68 KB

🖥️ Setup

This code was tested on Ubuntu 22.04 LTS, RTX A5000, CUDA 12.1.

Clone the repository:

git clone https://github.com/AndreeaDogaru/Gen3DSR
cd Gen3DSR
EXT_DIR=$(pwd)/external

For a successful installation, the locally installed CUDA toolkit should match with the one in the PyTorch binaries. If CUDA 12.1 is not available, we recommend downloading the corresponding runfile from here and installing the toolkit:

./runfile.run --silent --toolkit --toolkitpath=/usr/local/cuda-12.1/

Set environment variables to avoid CUDA conflicts:

export CUDA_HOME=/usr/local/cuda-12.1/
export CUDA_TOOLKIT_ROOT_DIR=$CUDA_HOME
export LD_LIBRARY_PATH="$CUDA_HOME/extras/CUPTI/lib64"  
export LIBRARY_PATH=$CUDA_HOME/lib64  
export LD_LIBRARY_PATH=$CUDA_HOME/lib64  
export CFLAGS="-I$CUDA_HOME/include $CFLAGS"

It is recommended to use a virtual environment with Python 3.10. Install requirements using pip:

pip install torch==2.2.2 torchvision==0.17.2 --index-url https://download.pytorch.org/whl/cu121
pip install -r requirements.txt

Install additional dependencies:

pip install git+https://github.com/jinlinyi/PerspectiveFields.git
pip install git+https://github.com/NVlabs/nvdiffrast
pip install git+https://github.com/ashawkey/kiuikit 
pip install https://github.com/AndreeaDogaru/mmcv/releases/download/v2.1.0/mmcv-2.1.0-cp310-cp310-linux_x86_64.whl
pip install $EXT_DIR/dreamgaussian/diff-gaussian-rasterization 
pip install $EXT_DIR/dreamgaussian/simple-knn 
python3 -m pip install -v -U xformers==0.0.25.post1 --index-url https://download.pytorch.org/whl/cu121
pip install git+https://github.com/open-mmlab/mmdetection.git 
pip install natten==0.15.1+torch220cu121 -f https://shi-labs.com/natten/wheels/ 
python -m pip install -e $EXT_DIR/detectron2 

cd $EXT_DIR/detectron2/projects/CropFormer/entity_api/PythonAPI
python3 setup.py build_ext --inplace
cd $EXT_DIR/detectron2/projects/CropFormer/mask2former/modeling/pixel_decoder/ops
python3 setup.py build install

You can download the checkpoints either manually by placing them in the specified directory or automatically using the included script. Optional: as all the checkpoints require about 30GB of local storage, you can selectively only download the ones for the models you plan to use. For the model provided by Adobe_EntitySeg, you must first accept the repository's terms and then use a Hugging Face token to download it.

cd $EXT_DIR/checkpoints
./download.sh <HF_TOKEN>
cd $EXT_DIR/..