GPU RTX3080 Pytorch error

# RTX 3080 GPU로 돌리면 마치 cuda가 지원안되는 것처럼 Torch 에러나는 경우있음

# 기존 설치 버전
# - torch 1.7.0, torchvision 0.8.1, cuda 10.2
# → ERR

 

# GPU RTX 3080 Error

• cuda install

wget https://developer.download.nvidia.com/compute/cuda/11.2.2/local_installers/cuda_11.2.2_460.32.03_linux.run
sudo sh cuda_11.2.2_460.32.03_linux.run
# only CUDA toolkit install

• upload cudnn in dms/etc to RTX3080 server and unzip it.

tar -xvf cudnn-11.2-linux-x64-v8.1.1.33

• copy cudnn related files to the place where cuda toolkit is installed

cp ./cuda/include/* /usr/local/cuda-11.2/include
cp -P ./cuda/lib64/* /usr/local/cuda-11.2/lib64
chmod a+r /usr/local/cuda-11.2/lib64/libcudnn*

• pytorch install

pip install torch==1.7.1+cu110 torchvision==0.8.2+cu110 torchaudio===0.7.2 -f https://download.pytorch.org/whl/torch_stable.html

• fvcore install

pip install fvcore==0.1.3.post20210204

• iopath install

pip install iopath==0.1.3

• detectron2 setup

python setup.py build develop