1. Make a U disk for the ubuntu system
1.1, download the ubuntu mirror: http://mirrors.163.com/ubuntu-releases/ download the required version
1.2 , Decompress the downloaded ubuntu image to a USB flash drive, and then make a USB flash drive for the ubuntu system.
2. Use the disk splitter that comes with win10 to split a part of the memory for ubuntu use
2.1, right click on’My Computer’ ->’Manage ‘->’Disk Management’
If you want to install ubuntu on Disk F, right-click Disk F ->’Compressed Disk’, and select the space required by ubuntu.
3. Install ubuntu
Plug the USB disk into the computer, restart the computer, enter the bios, select the USB disk to start, and select install ubuntu to start the installation of ubuntu. You need to pay attention to the installation process, choose ubuntu and win10 to coexist, and then it will be automatically installed in the space we just compressed.
4. Graphics card driver installation
4.1. Download the driver: https://www.nvidia.cn/Download/index.aspx?lang=cn Choose the appropriate graphics card drive.
4.2. Uninstall other graphics card drivers: sudo apt-get remove –purge nvidia*
4.3, disable nouveau:
sudo gedit /etc/modprobe.d/blacklist.conf
Add at the end of the text :
blacklist nouveau
option nouveau modeset=0
Run: sudo update-initramfs -u
Restart the computer
Run: lsmod | grep nouveau //If there is no output, then It shows that nouveau is successfully disabled.
4.4. Install the driver
Place the driver we downloaded in a directory with no Chinese path (because we can’t enter Chinese after a while), for example /home
Run: sudo service lightdm stop // Close the graphical interface
cd /home
sudo sh ./
reboot //Restart the computer and the installation is complete
5, install CUDA
5.1, download CUDA
https://developer.nvidia.com/ cuda-90-download-archive
Choose the version that suits your environment.
sudo sh ./
5.2. Option selection during installation: 
5.3. Modify environment variables:
sudo vim ~/.bashrc
Add at the end of the text:
export CUDA_HOME=/usr/local/cuda
export LD_LIBRARY_PATH=${CUDA_HOME}/lib64
export PATH=${CUDA_HOME}/bin:${PATH}
Exit and run: source ~/.bashrc //Enable the environment variables to take effect
Run nvcc -V //Test whether CUDA is installed successfully: If it displays as follows, then it succeeds:
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2018 NVIDIA Corporation
Built on Sat_Aug_25_21:08:01_CDT_2018
Cuda compilation tools, release 9.0, V9.0.130
6, cudnn installation:
https://developer.nvidia.com/rdp/cudnn-download //You need to register, and you can download it after logging in.
Download the cudnn corresponding to the cuda version.
After downloading the compressed package of cudnn, unzip it to get the cuda folder.
sudo cp cuda/lib64/ /usr/local/cuda/lib64/sudo cp cuda/include/ /usr/local/cuda/include/