Ubuntu深度学习环境配置

1. 显卡驱动

• 下载对应驱动 https://www.nvidia.cn/Download/index.aspx?lang=cn
• 卸载原有驱动 sudo apt-get remove --purge nvidia*
• 禁用nouveau(nouveau是通用的驱动程序) sudo gedit /etc/modprobe.d/blacklist.conf 或者(blacklist-nouveau.conf)
• 在打开的blacklist.conf末尾添加

blacklist nouveau
options nouveau modeset=0

• 保存后，在终端输入如下，更新下

sudo update-initramfs –u

• 等待结束后重启电脑，重启后在终端输入如下，没有任何输出表示屏蔽成功 lsmod | grep nouveau

• 安装lightdm，lightdm是显示管理器

sudo apt-get install lightdm

• 安装驱动：在终端输入sudo telinit 3

• 要回到图形化界面输入sudo telinit 5

• 输入你的用户名和密码登录。cd进入到你驱动下载的目录，输入命令sudo chmod 777 NVIDIAxxxx.run

• sudo ./NVIDIA-Linux-x86_64-430.26.run

nvidia-smi

2. GCC

sudo apt update
sudo apt install build-essential
gcc --version

3. CUDA

cuda下载地址， https://developer.nvidia.com/cuda-toolkit-archive

wget https://developer.download.nvidia.com/compute/cuda/11.3.0/local_installers/cuda_11.3.0_465.19.01_linux.run
sudo sh cuda_11.3.0_465.19.01_linux.run

nvcc -V

cp -fr include/* /usr/local/cuda-11.3/include/
cp -fr lib64/* /usr/local/cuda-11.3/lib64/

cuda卸载

cd /usr/local/cuda6.5/bin/
sudo ./uninstall_****
sudo apt-get purge --auto-remove nvidia-cuda-toolkit

打开~/.bashrc

export PATH=$PATH:/opt/cuda/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/cuda/lib:/opt/cuda/lib64 

4. Anaconda

官网地址， https://www.anaconda.com/products/individual#linux

wget https://repo.anaconda.com/archive/Anaconda3-2020.11-Linux-x86_64.sh

5. 其他python相关库

安装pytorch

conda install pytorch=1.7.0 torchvision torchaudio cudatoolkit=10.2 -c pytorch