search

zhihao-lin / neurmips

Pytorch implementation of paper: "NeurMiPs: Neural Mixture of Planar Experts for View Synthesis"
MIT License
113 stars 11 forks source link

Any luck running this on a 3000 series GPU? #1

Open reynoldscem opened 2 years ago

reynoldscem commented 2 years ago

It appears the version of cuda toolkit and maybe some other stuff set in the environment.yml isn't compatible with the sm_86 compute arch. Have you tried running this on nVidia 3000 series cards and can you advise how one might successfully install / compile / run the code on them?

haoyuhsu commented 2 years ago

For running this on NVIDIA 3000 series cards, you can follow the instructions below for environment installation & CUDA code compile. (note that this has been tested on RTX 3090 card successfully)

Install PyTorch

conda create -n neurmips python=3.8 conda activate neurmips conda install -c pytorch pytorch=1.8.0 torchvision cudatoolkit=11.1 conda install -c conda-forge -c fvcore -c iopath fvcore iopath

Option 1: Install Pytorch3D

conda install pytorch3d -c pytorch3d

Option 2: Install Pytorch3D from Github (recommended) (if option 1 do NOT work)

pip install "git+https://github.com/facebookresearch/pytorch3d.git"

Install Hydra

conda install -c conda-forge hydra-core

Install associate packages

pip install vedo pip install scikit-image

change nvcc to installed version (CUDA 11.1)

export PATH=/usr/local/cuda-11.1/bin${PATH:+:${PATH}} export LD_LIBRARY_PATH=/usr/local/cuda-11.1/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

compile CUDA code

cd cuda python setup.py develop

reynoldscem commented 2 years ago

Thank you, I will try that out.