DockerDL

Deep Learning Docker Image

Don't waste time on setting up a deep learning environment while you can get a deep learning environment with everything pre-installed.

List of Packages installed

• conda
• Jupyter lab
• Matplotlib
• NLTK
• Numpy
• Pandas
• Plotly
• PyTorch
• Scikit-Learn
• Seaborn
• TensorFlow
• zellij

Image variants and tags

Variant	Tag	Conda	PyTorch	TensorFlow	Image size
Conda	conda	✔️	❌	❌
Tensorflow	tf	❌	❌	✔️
PyTorch	torch	❌	✔️	❌
PyTorch + Tensorflow	tf-torch, latest	❌	✔️	✔️
PyTorch + Tensorflow + Conda	tf-torch-conda	✔️	✔️	✔️

You can see the full list of tags https://hub.docker.com/r/matifali/dockerdl/tags.

Requirements

1. Docker
2. nvidia-container-toolkit ¹
3. Linux, or Windows with WSL2

Fast Start

docker run --gpus all --rm -it -h dockerdl matifali/dockerdl bash

JupyterLab server without conda

docker run --gpus all --rm -it -h dockerdl -p 8888:8888 matifali/dockerdl jupyter lab --no-browser --port 8888 --ServerApp.token='' --ip='*'

Connect by opening http://localhost:8888 in your browser.

Customize the image

Clone the repo

git clone https://github.com/matifali/dockerdl.git

Add or delete packages

Modify the corresponding [Dockerfile] to add or delete packages.

Note

You may have to rebuild the dockerdl-base if you are building a custom image and then use it as a base image. See Build section.

Build

The following --build-arg are available for the dockerdl-base image.

Argument	Description	Default	Possible Values
USERNAME	User name	coder	Any string or $USER
USERID	User ID	1000	$(id -u $USER)
GROUPID	Group ID	1000	$(id -g $USER)
CUDA_VER	CUDA version	12.4.1
UBUNTU_VER	Ubuntu version	22.04	22.04, 20.04, 18.04

Warning

Not all combinations of --build-arg are tested.

Step 1

Build the base image

docker build -t dockerdl-base:latest --build-arg USERNAME=coder --build-arg CUDA_VER=12.4.1 --build-arg UBUNTU_VER=22.04 -f base.Dockerfile .

Step 2

Build the image you want with the base image as the base image.

docker build -t dockerdl:tf --build-arg TF_VERSION=2.12.0 -f tf.Dockerfile .

or

docker build -t dockerdl:torch --build-arg -f torch.Dockerfile .

How to connect

VS Code

1. Install vscode.
2. Install the following extensions:
   1. Docker.
   2. Python.
   3. Remote Development.
3. Follow the instructions here.

Coder

1. Install Coder. (https://github.com/coder/coder).
2. Use deeplearning template which references these images (https://github.com/matifali/coder-templates/tree/main/deeplearning).

JetBrains PyCharm Professional

Follow the instructions here.

Issues

If you find any issue please feel free to create an issue and submit a PR.

Support

• Give a star (⭐) if using this has helped you.
• 

References

Footnotes

1. This image is based on nvidia/cuda and uses nvidia-container-toolkit to access the GPU. ↩