Ubuntu20.04安装NVIDIA显卡驱动、CUDA、CUDNN、TensorRT

一、Nidia显卡驱动(apt install)

1. 禁用自带驱动

sudo gedit /etc/modprobe.d/blacklist_nouveau.conf

# 在/etc/modprobe.d/blacklist_nouveau.conf中加入下面两行命令
blacklist nouveau
options nouveau modeset=0

2. 卸载显卡驱动

sudo apt-get --purge remove nvidia-*

3. 查看系统中驱动版本，选择推荐的驱动版本进行安装：如安装535版本驱动命令如下

ubuntu-drivers devices
sudo apt install nvidia-driver-535

4. 重启电脑

reboot

5. 验证

nvidia-smi

二、CUDA(`run`文件安装)

1. 下载cuda安装包：https://developer.nvidia.com/cuda-toolkit-archive
2. 安装

sodo sh cuda_xxxxx.run

3. 选择不安装驱动

4. 设置环境变量

export CUDA_HOME=/usr/local/cuda-10.2
export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:${CUDA_HOME}/lib64
export PATH=${CUDA_HOME}/bin:${PATH}

5. 验证

# 查看版本信息
nvcc -V

# 编译测试
cd /usr/local/cuda-10.2/samples/1_Utilities/deviceQuery
make
./deviceQuery

三、CUDNN

1. 下载CUDNN安装包：https://developer.nvidia.com/rdp/cudnn-archive
2. 解压

tar -xvf cudnn-linux-x86_64-8.x.x.x.tar.xz

3. 安装

sudo cp include/* /usr/local/cuda/include/
sudo cp lib64/* /usr/local/cuda/lib64/
sudo chmod a+r /usr/local/cuda/include/cudnn*.h usr/local/cuda/lib64/libcudnn*

4. 验证

cat /usr/local/cuda/include/cudnn.h | grep cudnn

四、TensorRT

参考资料：
https://docs.nvidia.com/deeplearning/tensorrt/install-guide/index.html#installing-tar
https://zhuanlan.zhihu.com/p/392143346
https://blog.51cto.com/u_12870633/6149817
https://blog.youkuaiyun.com/u010420283/article/details/134634995?spm=1001.2014.3001.5506

1. 下载TensorRT安装包：https://developer.nvidia.com/tensorrt/download
2. 解压

version="8.x.x.x"
arch=$(uname -m)
cuda="cuda-x.x"
tar -xzvf TensorRT-${version}.Linux.${arch}-gnu.${cuda}.tar.gz

3. 移动到指定文件夹

mv TensorRT-${version} /usr/local

```bash

```
3. 添加环境变量

sudo gedit ~/.bashrc


export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/TensorRT-${version}/lib
export C_INCLUDE_PATH=$C_INCLUDE_PATH:/usr/local/TensorRT-${version}/include
export CPLUS_INCLUDE_PATH=$CPLUS_INCLUDE_PATH:/usr/local/TensorRT-${version}/include


source ~/.bashrc

4. 验证

cd /usr/local/TensorRT-${version}/samples/sampleOnnxMNIST
sudo make
/usr/local/TensorRT-${version}/bin/sample_onnx_mnist

5. 验证报错

sudo rmmod nvidia_uvm
sudo modprobe nvidia_uvm