Skip to content

Latest commit

 

History

History
81 lines (62 loc) · 2.75 KB

install.md

File metadata and controls

81 lines (62 loc) · 2.75 KB

Installation

sudo apt install libsparsehash-dev # you can try to install sparsehash with conda if you don't have sudo privileges.

1.install enviroment.

if your gcc version in your system >=7.2

conda env create -f requirements/environment.yaml
conda activate deep3dmap2

else you need to install gcc first.

conda env create deep3dmap2
conda activate deep3dmap2
conda install -c moussi gcc_impl_linux-64
conda install -c moussi gcc_linux-64
conda install -c moussi gxx_impl_linux-64
conda install -c moussi gxx_linux-64
conda env update -f requirements/environment.yaml

note:pytorch is include in requirements/environment.yaml, you can take out it and install it by yourself, we recommend the version >=1.6, and you can choose other cuda version according to your machine.

2.install python packages.

pip install -r requirements/requirements.txt

[FAQ on installation torchsparse]

  • AttributeError: module 'torchsparse_backend' has no attribute 'hash_forward'

  • Clone torchsparse to a local directory. If you have done that, recompile and install torchsparse after removing the build folder.

  • No sudo privileges to install libsparsehash-dev

  • Install sparsehash in conda (included in environment.yaml) and run export CPLUS_INCLUDE_PATH=$CONDA_PREFIX/include before installing torchsparse.

  • if you want to rebuild torchsparse, please rm build folder and .so files

  • For other problems, you can also refer to the FAQ in torchsparse.

[FAQ on import tinycudann] - 'undefined symbol:_ZN2at4_ops19empty_memory_format4callEN3c108ArrayRefIlEENS2_8optionalINS2_10ScalarTypeEEENS5_INS2_6LayoutEEENS5_INS2_6DeviceEEENS5_IbEENS5_INS2_12MemoryFormatEEE' - Install version not match between tinycudann and pytorch, uninstall and reinstall them.

3.install pnpmodules

first specify cuda path with the same version for pytorch

export PATH="/usr/local/cuda-11.xxx/bin:$PATH"
export LD_LIBRARY_PATH="/usr/local/cuda-11.xxx/lib64:$LD_LIBRARY_PATH"
  • stylegan2
cd pnpmodules/stylegan2/stylegan2-pytorch/op/
python setup.py install
  • neural_renderer
cd pnpmodules/neural_renderer
python setup.py install
  • "ImportError: libc10.so: cannot open shared object file: No such file or directory"
  • libc10.so is based on torch, import torch first

4.install cuda build operators.

To use the cuda ops, you need to execute the build operation:

 export CUDA_HOME=$CUDA_PATH_Vxx.x
 export PATH=$CUDA_HOME/bin:$PATH
 export LD_LIBRARY_PATH=$CUDA_HOME/lib64:$LD_LIBRARY_PATH
 export TORCH_CUDA_ARCH_LIST="3.5 3.7 5.0 5.2 6.0 6.1 7.0 7.5 8.0"
 WITH_OPS=1 python setup.py build_ext