Deploy Base Image

Learn to deploy Docker images in Code Ocean.

You can deploy Docker images as starter environments for compute capsules. The starter environments are also called base images and will be available for users in their capsule's environment editor. The starter environments may include different machine languages, versions of languages and IDEs, software packages, etc., so users can start customizing their ideal compute environments efficiently.

You can deploy a recommended base image or switch tags to build a custom image.

Click on +Add New Base Image and the ‘New Base Image’ dialog will appear with the ‘Recommended’ view.

Deploy Recommend Base Image

We currently provide three base images to deploy. You can select the images and click on the Deploy button to deploy the recommended base image.

Deploy Custom Image

Complete the fields and click Deploy to deploy the image (see an explanation about each field below).

Fill fields from existing image

The system keeps a list of images deployed in each VPC account. Clicking on it will show all the deployed image names.

Name (required, editable)

The name of the image. Any name that reflects the language or purpose of the image is suitable.

Docker Image (required)

The URL of the image to pull from. Any public Docker image URL or any base image from the Code Ocean Docker registry starting with registry.codeocean.com works. For private image URLs, contact [email protected]

Version (required, editable)

The version of the image so you can keep track of future updates. Different versions of images under the same name will be grouped together in the Environment Editor.

Language (required, editable)

The programming language tag for this image. The language tag can be used for filtering in the Environment manager.

Description (optional)

The description of this base image so end users can understand what the image is about.

Machine Type (optional)

Standard: with CPU only, GPU: with GPU.

Tags (optional)

Keywords relevant to the image such as language, version, etc., so it's easier to search the image.

• Name + version combination must be unique

• The docker image source of the new image must be unique

The default user in the container from the new image needs to be the root user for it to access the folder in the cloud workstation and the computation. Keep this in mind if pulling from a public docker image. If you are creating the docker file, please add this line: USER 0:0 ( which means set the user inside the container to root).

Deploying Docker Images Examples

Below are the examples of completed forms from the recommend base image:

Python

This is an example of a Python image that runs on CPU-only instances:

R

Python with GPU support

Tracking Deployment Status

After you click on Deploy, you can track the deployment status in the images list.