dl-dev (deep learning development)

This docker image support a bunch of deep learning platform and easy to development. The developer is easy to debug and get around with all the dependencies within the container.

Support deep-learning platform list:

Platform Name	Version	Language
Caffe	1.0.0	C++, Python
Tensorflow	1.12.0	Python

Prerequesties

• GNU/Linux x86_64 with kernel version > 3.10
• Docker >= 1.12
• NVIDIA GPU with Architecture > Fermi (2.1)
• NVIDIA drivers ~= 361.93 (untested on older versions)

Get started(setup)

• First, you should download docker yourself :). Check the following infomation:

  • For Mac.
  • For Ubuntu.
  • For Windows.

  However, I highly recommend you use unix-like system since docker is well-supported on such OS.

• After successfully installed docker, run command sudo usermod -aG docker ${USER} to add yourself into user-group within system wide.

• Reboot by invoking command sudo reboot.

• Change directory into cd $DIR-TO-DL-DEV.

• Install nvidia-docker for GPU usage, run command ./scripts/setup_nvidia_docker.sh.

How to use docker container and the dl-dev image

You can learn some basic docker information at docker's website here. I recommend you exercise the tutorial around.

• To start a container named dl-dev, you can do it by running the command ./scripts/start.sh. Note: this process is only needed if you have not start that container on the host before. It is only needed if you first start or reboot your computer.

  as the above image said, container name is dl-dev. It will pull the newest image from docker hub and start the container.

• Everytime you want to dive into the container and start programming, just run ./scripts/into.sh. It should looks like this:

  

About the docker container

• Within the docker container(dl-dev for this specific project), you are login as user "dl" and the passwd is already set to "abcd".

• The directory of the workspace looks like this:

  

• The 'Downloads' directory is mount to your host machine's Downloads directory, so you can store your data outside docker.

• The 'Codes' directory is designed to store your code and project, this directory is invisible outside docker.

• In terms of caffe, unpooling layer is added and you can find the repo here: caffe

• zsh: Use oh-my-zsh theme for shell, and support zsh functionalities. Kindly check this websit.

• VS Code： You can issue code in zsh, then the container will open IDE in host machine.

• tmux: The configuration of tmux is as default, you can check useful information in this page.

To build the docker image

To build customized docker iamge yourself, please check build reference. I do not recommand do this process unless you really need it.

Issues

Please contact [email protected] for any issue. To add support, please also email me.