RTX3080/RTX3090驱动安装CUDA11.1+CUDNN8.0.4.30+pytorch源码编译

一、nvidia官网下载nvidia-driver455.23以上，cuda11.1,cudnn-v8.0.4.30

确保nvidia-driver、cuda11.1、cudnn8.0.4.30安装成功

• NVIDIA-Linux-x86_64-455.23.04.run
• cuda_11.1.0_455.23.05_linux.run
• cudnn-11.1-linux-x64-v8.0.4.30.tgz/8.0.5.39
• 上述文件百度网盘下载链接nvidia-cuda-cudnn 密码: paat，nvidia-driver本人已经更新到455.45.01，cudnn:8.0.5.39
  
  

二、pytorch+torchvision源码编译

• conda create -n pytorch python=3.7

• pip install numpy ninja pyyaml mkl mkl-include setuptools cmake cffi typing_extensions future six requests dataclasses

• git clone --recursive https://github.com/pytorch/pytorch
  cd pytorch
  git submodule sync
  git submodule update --init --recursive
  

• export CMAKE_PREFIX_PATH=${CONDA_PREFIX:-"$(dirname $(which conda))/../"}
  python setup.py install
  

• 成功之后重启终端

  python
  >>> import torch
  >>> torch.__version__
  '1.8.0a0+3d421b3'
  >>> torch.cuda.is_available()
  True
  
  

• git clone --recursive https://github.com/pytorch/vision.git

• cd vision/
  python setup.py install
  

• 重启终端

• >>> import torchvision
  >>> torchvision.__version__
  '0.8.0a0+a9c78f1'
  >>> 
  

三、tensorflow安装

• pip install tf-nightly-gpu
• >>> tf.__version__ '2.4.0-dev20201014'