Docker for A.I. Researcher

"Docker for A.I. Researcher" is a series of Shell script that

• Allows you to quickly set up your deep learning research environment.
• Supports SSH! Remote work with laptop OUTSIDE of the box.
• Supports the next-generation pytorch library, pytorch-lightning.
• Includes fancy terminal setup with oh-my-zsh.

Demo

Docker: Try it on Binder

---

1. Requirements

• Nvidia GPU Driver Installation
• 10 minuites

---

2. What's inside?

├── src/
│   ├── Dockerfile
│   └── requirements.txt
├── .env
├── 1_1-install_nvidia_driver.sh
├── 1_2-terminal_setting.sh
├── 2-docker_install.sh
├── 3-docker_build.sh
├── 4-run_docker.sh
└── 5-setting_alias.sh

• 1_1-install_nvidia_driver.sh : Install NVIDIA Driver
• 1_2-terminal_setting.sh : Setting Terminal with zsh
• 2-docker_install.sh : Install Docker + NvidiaDocker2
• 3-docker_build.sh : Build docker image.
• 4-run_docker.sh : Run Docker Container.
• 5-setting_alias.sh : Optional. Setting alias for conviniences.

3. Quick Start

step 1. clone the repository

sudo apt-get install git
git clone https://github.com/eungbean/Docker-for-AI-Researcher
cd Docker-for-AI-Researcher

Step 2. Install NVIDIA Driver and Terminal setup.

./1_1-install_nvidia_driver.sh
./1_2-terminal_setting.sh

Install following packages.

• NVidia Driver
• Terminal tools
  • zsh
  • oh-my-zsh
  • zsh-syntax-highlighting
  • zsh-autosuggestions
  • neovim
  • spacevim
  • powerline font
• GPU Monitoring tools
  • gpustat
  • glances
• git
• openssh

Step 3. Install Docker

./2-docker_install.sh

exact same procedure from Nvidia Docker 2 Installation Guide.

Step 4. Configure Your Environment

1. configure bining folders, ports, etc.

.env

2. (Optional) ssh id/password.

• Initial root user id/password is root:root.
• If you want to change ssh id/password, change line below in src/Dockerfile.

RUN echo 'root:root' |chpasswd

• or, change id/password with passwd command.

3. (Optional) base docker image.

• If you want to change base images, change 1st line in src/Dockerfile.

FROM pytorchlightning/pytorch_lightning:base-cuda-py3.8-torch1.9

Step 5. Build and run your Docker Container

./3-docker_build.sh
./4-run_docker.sh

• SSH to container will set automatically.

Step 6. Post Installation

1) Send public key to container.

ssh-copy-id -p 10022 [email protected]

• Initial ssh id/pw is root/root.

2) ssh to docker container.

with ssh

ssh -p 10022 [email protected]

• if you assign 10022 port for ssh, then you can access to your container via 10022 port from your laptop.

docker exec

docker exec -it ${CONTAINER_NAME} /usr/bin/zsh

3) Change ssh password

(inside docker container)

passwd

You have to change password.