Deploying custom Docker images

Deeploy allows you to deploy custom Docker images. With the Deeploy python client CLI you can easily create your own custom Docker images for models, explainers and transformers. Make sure to follow the Deeploy python client installation instructions and then run the deeploy generate-template --help command.
The generate-template command initializes a folder with all files that you need in order to embed your own custom logic whilst adhering to the Deeploy best practices and API spec. Make sure to follow the steps in the README.md that is located in the created folder.

Reference the Docker image

To deploy a custom Docker image, you need to reference the image in your Repository using a reference.json file. This is further explained in Repository Requirements.

This is an example of a reference.json file for a custom Docker image:

# this reference is used in example repository:
# https://gitlab.com/deeploy-ml/sample-models/example-custom-image-hello-world/
{
    "reference": {
        "docker": {
            "image": "deeployml/example-custom-image-hello-world:0.3.0",
            "port": 8080
        }
    }
}

Parameter	Description	Example
image	The link to a Docker Image, including the tag	example/image:1.2.3
uri (optional)	The URI on which to request predictions	/model:predict
port	The port number to expose	8000

Adding Docker credentials

If the Docker image requires credentials to access, add the credentials through the Deeploy front-end, as explained in Docker image credentials.

Deploy the custom Docker image

Follow the steps outlined in Creating a Deployment. Make sure to select the custom docker option as your model, explainer or transformer framework.

Note

Absence of entrypoint in your custom Docker images may lead to issues. It is recommended to use ENTRYPOINT over CMD in your Dockerfile.

Environment variables

The ability to define environment variables is in development and currently only available through the Python Client or API.