Junjie Su

Junjie Su

he/his

An undergraduate from BUPT

Installing Ubuntu 20.04 for Development and Deep Learning

9 minute read

Published:

This guide walks you through installing Ubuntu Server 20.04 on a physical machine with a GPU, setting up a basic development environment with Docker and Docker Compose, and configuring CUDA for deep learning. These steps generally apply to other installation methods and Ubuntu versions (18.04-24.04), so you can follow this whether you’re installing the Desktop or Server version, on a physical machine or a virtual machine. If you’re using a VM, you can skip the USB drive creation.

1. Create a Bootable USB Drive

First, you’ll need a USB drive (at least 4GB, 16GB recommended) to burn the Ubuntu image onto.

Important: Back up everything on the USB drive as it will be formatted.

Next, download the Ubuntu 20.04 Server image from:

  • ubuntu.com: Official site, usually the most stable and fastest (may require a proxy).
  • Tsinghua Mirror: A mirror site that can be used without a proxy, but might be slower.

Then, download a tool to create the bootable USB drive:

  • Windows: Rufus (refer to this guide - ensure you select the correct Ubuntu version).
  • macOS: balenaEtcher (user-friendly, just follow the on-screen instructions).

Once done, you should have a bootable USB drive.

2. Install Ubuntu from USB

  1. Plug the USB drive into your server, ensuring it’s connected to the motherboard.
  2. Start the server and press F12 (or F2, ESC, check the boot screen) repeatedly to enter the BIOS.
  3. In the BIOS, change the boot order to prioritize the USB drive, or directly select the USB drive as the boot device.
  4. Save the changes and exit the BIOS. The Ubuntu installation process should begin.

2.1 Installation Details

Follow the steps in this video for a visual guide. Customize settings like disk partitioning as needed.

Tips:

  • Internal Network: If installing on an internal network requiring a specific server address, ensure you have the static IP, subnet mask, and gateway IP.
  • APT Sources and Updates: Avoid configuring APT sources or updating system packages during the initial installation. Stick with the default sources and skip the update step. Changing sources or updating with network issues can lead to installation failures and the need to reinstall.

3. Configure System Network

This is crucial for remote server access on physical machines. VMs are simpler to configure via their tools, but understanding these concepts is important for later package installations.

3.1 Configure System Gateway

We recommend using netplan, which comes pre-installed on Ubuntu Server. The configuration file for Ubuntu Server 22.04 is typically at /etc/netplan/00-installer-config.yaml.

sudo vim /etc/netplan/00-installer-config.yaml

Changes here are permanent and applied on system boot. For temporary configuration, use:

sudo route add -net <target_network> netmask <subnet_mask> gw <gateway_IP>

Configuration varies by system and netplan version. Here’s a sample:

network:
  version: 2
  renderer: networkd
  ethernets:
    eth0:
      dhcp4: no
      addresses: [192.168.1.100/24]
      routes:
        - to: default # Default route for external network
          via: 192.168.1.1 # Gateway IP
        - to: 10.0.0.0/8 # Specific network for internal machines
          via: 192.168.1.2 # Gateway IP
      nameservers:
        addresses: [8.8.8.8, 8.8.4.4]

Apply the configuration:

sudo netplan try  # Test configuration (reverts on no confirmation)
# Or apply directly:
sudo netplan apply

Verify the configuration:

ip route show

Tip: If configuring the network remotely via SSH, use sudo netplan try to test your changes. If you don’t confirm within the timeout, the configuration will revert, preventing you from losing connection due to errors.

3.2 Configure Basic Tools

Once networking is set up, you can SSH into your server from your local machine:

ssh <username>@<hostname_or_IP>

3.2.1 APT

First, configure APT sources, especially if you’re in China. Edit the sources list:

sudo vim /etc/apt/sources.list

Replace the contents with a mirror like Aliyun:

deb [http://mirrors.aliyun.com/ubuntu/](http://mirrors.aliyun.com/ubuntu/) focal main restricted universe multiverse
deb [http://mirrors.aliyun.com/ubuntu/](http://mirrors.aliyun.com/ubuntu/) focal-security main restricted universe multiverse
deb [http://mirrors.aliyun.com/ubuntu/](http://mirrors.aliyun.com/ubuntu/) focal-updates main restricted universe multiverse
deb [http://mirrors.aliyun.com/ubuntu/](http://mirrors.aliyun.com/ubuntu/) focal-backports main restricted universe multiverse
deb [http://mirrors.aliyun.com/ubuntu/](http://mirrors.aliyun.com/ubuntu/) focal-proposed main restricted universe multiverse
deb [arch=amd64] [http://mirrors.aliyun.com/docker-ce/linux/ubuntu](http://mirrors.aliyun.com/docker-ce/linux/ubuntu) focal stable
# deb-src [arch=amd64] [http://mirrors.aliyun.com/docker-ce/linux/ubuntu](http://mirrors.aliyun.com/docker-ce/linux/ubuntu) focal stable

Save and update the APT sources and upgrade existing packages:

sudo apt update && sudo apt upgrade

This might take some time. Then, install essential tools:

sudo apt install build-essential  # Includes GCC and other common tools
sudo apt install docker docker-compose  # For backend deployment

3.2.2 Docker

Docker is a vital tool for production environments. Here’s a more detailed setup:

Reference: This article provides a good overview of Docker and Docker Compose installation.

3.2.2.1 Docker Installation

While you can install Docker with apt, this method forgoes easier mirror configuration. We recommend this approach:

Install Docker dependencies:

sudo apt-get install ca-certificates curl gnupg lsb-release

Add Docker’s official GPG key:

curl -fsSL [http://mirrors.aliyun.com/docker-ce/linux/ubuntu/gpg](http://mirrors.aliyun.com/docker-ce/linux/ubuntu/gpg) | sudo apt-key add -

Configure the Docker mirror (Aliyun):

sudo add-apt-repository "deb [arch=amd64] [http://mirrors.aliyun.com/docker-ce/linux/ubuntu](http://mirrors.aliyun.com/docker-ce/linux/ubuntu) $(lsb_release -cs) stable"

Now, install Docker CE:

sudo apt-get install docker-ce docker-ce-cli containerd.io

Docker is now installed. To avoid using sudo with Docker commands, add your user to the docker group:

sudo usermod -aG docker $USER

You might need to log out and back in for this to take effect. Verify the installation:

sudo systemctl start docker
sudo systemctl status docker
sudo docker run hello-world

If your network is working, the test image should run successfully. Enable Docker to start on boot:

sudo systemctl enable docker

Optional tools:

sudo apt-get -y install apt-transport-https ca-certificates curl software-properties-common

Restart Docker after installing new tools or changing configurations:

sudo systemctl restart docker

To change Docker settings like mirrors, edit /etc/docker/daemon.json (requires sudo). After editing, reload the daemon and restart Docker:

sudo systemctl reload docker
sudo systemctl restart docker

3.2.2.2 Docker Compose Installation

Install Docker Compose:

# GitHub: [https://github.com/docker/compose/releases/tag/v2.20.2](https://github.com/docker/compose/releases/tag/v2.20.2)
# China Mirror: [https://gitee.com/smilezgy/compose/releases/tag/v2.20.2](https://gitee.com/smilezgy/compose/releases/tag/v2.20.2)
sudo curl -SL \
[https://github.com/docker/compose/releases/download/v2.20.2/docker-compose-linux-x86_64](https://github.com/docker/compose/releases/download/v2.20.2/docker-compose-linux-x86_64) \
-o /usr/local/bin/docker-compose

# Or download manually, upload to the server, and run:
# In the same directory as docker-compose-linux-x86_64
sudo cp docker-compose-linux-x86_64 /usr/local/bin/docker-compose

# Make it executable
sudo chmod +x /usr/local/bin/docker-compose

Verify the installation:

docker-compose --version

Docker and Docker Compose are now set up. If you encounter download timeouts, configure Docker mirrors in /etc/docker/daemon.json:

{
    "registry-mirrors": [
        "[https://registry.docker-cn.com](https://registry.docker-cn.com)",
        "[https://mirror.ccs.tencentyun.com](https://mirror.ccs.tencentyun.com)",
        "[https://docker.mirrors.ustc.edu.cn](https://docker.mirrors.ustc.edu.cn)",
        "[https://docker.m.daocloud.io](https://docker.m.daocloud.io)",
        "[https://docker.imgdb.de](https://docker.imgdb.de)",
        "[https://docker-0.unsee.tech](https://docker-0.unsee.tech)",
        "[https://docker.hlmirror.com](https://docker.hlmirror.com)",
        "[https://docker.1ms.run](https://docker.1ms.run)",
        "[https://func.ink](https://func.ink)",
        "[https://lispy.org](https://lispy.org)",
        "[https://docker.xiaogenban1993.com](https://docker.xiaogenban1993.com)"
    ]
}

Test a few mirrors to see which work best for you.

4. Install NVIDIA Driver and CUDA

4.1 NVIDIA Driver

For a 1080Ti (example GPU):

  1. Go to the NVIDIA Drivers website.
  2. Select your GPU model and operating system, then search for the latest driver.
  3. Download the .run file.
  4. Make it executable and run it:
chmod +x NVIDIA-Linux-x86_64-*.run
sudo ./NVIDIA-Linux-x86_64-*.run

Installation takes time, and a reboot is usually required. Before rebooting, double-check your netplan configuration to avoid losing connection. Verify the installation:

nvidia-smi

Tip: This command shows GPU status. For monitoring during deep learning:

watch -n 1 nvidia-smi

This updates the GPU status every second, useful for tracking utilization.

4.2 CUDA Installation

For CUDA 12.8 (example version): The latest version might not support older GPUs. Check your GPU compatibility before proceeding.

  1. Search online for “cuda [your version]” (e.g., “cuda 12.8”).
  2. Go to the NVIDIA CUDA Toolkit download archive.
  3. Select your OS, architecture, distribution, version, and installer type (deb (local) recommended).
  4. Follow the instructions provided on the NVIDIA website. For the example CUDA 12.8 and Ubuntu 20.04:
wget [https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin](https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin)
sudo mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget [https://developer.download.nvidia.com/compute/cuda/12.8.1/local_installers/cuda-repo-ubuntu2004-12-8-local_12.8.1-570.124.06-1_amd64.deb](https://developer.download.nvidia.com/compute/cuda/12.8.1/local_installers/cuda-repo-ubuntu2004-12-8-local_12.8.1-570.124.06-1_amd64.deb)
sudo dpkg -i cuda-repo-ubuntu2004-12-8-local_12.8.1-570.124.06-1_amd64.deb
sudo cp /var/cuda-repo-ubuntu2004-12-8-local/cuda-*-keyring.gpg /usr/share/keyrings/
sudo apt-get update
sudo apt-get -y install cuda-toolkit-12-8

The installation will take a while.

5. Deep Learning Environment

5.1 Miniconda

  1. Go to the Miniconda installation page and get the latest Linux x64 installation instructions. As of May 14, 2025:
mkdir -p ~/miniconda3
wget [https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh](https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh) -O ~/miniconda3/miniconda.sh
bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
rm ~/miniconda3/miniconda.sh

  1. The installer will ask if you want to initialize Conda. We recommend saying yes for automatic ~/.bashrc configuration.

  2. After installation, your terminal prompt should show (base):

(base) user@hostname:~$
  1. Configure pip mirrors (Aliyun example):
mkdir ~/.pip
cd ~/.pip
vim pip.conf

Add the following content:

[global]
index-url = [https://mirrors.aliyun.com/pypi/simple/](https://mirrors.aliyun.com/pypi/simple/)

[install]
trusted-host = mirrors.aliyun.com
  1. Create a virtual environment (e.g., for PyTorch):
conda create -n torch python=3.9 -y
  1. Activate the environment and install PyTorch and other dependencies:
conda activate torch
pip install --upgrade pip
pip install torch torchvision torchaudio jupyter
  1. Verify the PyTorch and CUDA setup:
(torch) user@hostname:~$ python
Python 3.9.21 (main, Dec 11 2024, 16:24:11)
[GCC 11.2.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> torch.cuda.is_available()
True

If True is printed, your PyTorch and CUDA environment is configured correctly.

6. Conclusion

By following these steps, you’ve successfully installed Ubuntu Server 20.04 on a GPU-equipped machine and set up the necessary environment for backend development and deep learning. Each step, from creating the bootable drive to configuring the software, is crucial. This guide should help you overcome common installation and configuration challenges, regardless of your Ubuntu version or whether you’re using a physical or virtual machine. We hope this helps you quickly establish your ideal development and deep learning environment.

You May Also Enjoy

A Day in the Server Room: My Sysadmin Adventure

3 minute read

Published:

As a junior undergraduate, I’ve been given the chance to work in my mentor’s lab, a head start on research thanks to my confirmed spot in their graduate program after graduation. My days are usually spent coding algorithms, but on April 24, I faced a different kind of challenge. After an 8 AM class, I took the campus shuttle to the lab to deploy an AI backend I’d been developing. My mentor handed me an ancient server—its password long forgotten—and asked me to reinstall it from scratch. Just like that, I went from algo engineer to makeshift sysadmin.

Jwt Token

11 minute read

Published:

I’ve been busy working on the rebuttal for UAI 2025 these days. As a result, I haven’t written any blogs, nor have I developed any applications. Yesterday, our group held a meeting to discuss the new features of our application on Table Tennis. Unfortunately, I missed it and forgot to inform my mentor. Fortunately, F senpai covered for me. So, as soon as I finish the rebuttal, I offered to assist him with his graduation project as a way of repaying the favor.

Finding My Way

13 minute read

Published:

Today, I resolved to write blogs once more—or perhaps to truly begin. Three years ago, as a fledgling freshman at Beijing University of Posts and Telecommunications, I scribbled thoughts on momoyeyu.github.io. Those fragments, though, were mere whispers of notes, not the full-throated voice of a blog. So here I stand, on this cusp of March 31, 2025, claiming the craft anew, as if for the first time.