Docker Model Runner¶

Introduction¶

The Testcontainers module for DockerModelRunner.

Adding this module to your project dependencies¶

Please run the following command to add the DockerModelRunner module to your Go dependencies:

go get github.com/testcontainers/testcontainers-go/modules/dockermodelrunner

Usage example¶

Creating a DockerModelRunner container

ctx := context.Background()

const (
    modelNamespace = "ai"
    modelName      = "smollm2"
    modelTag       = "360M-Q4_K_M"
    fqModelName    = modelNamespace + "/" + modelName + ":" + modelTag
)

dmrCtr, err := dockermodelrunner.Run(
    ctx,
    dockermodelrunner.WithModel(fqModelName),
)
defer func() {
    if err := testcontainers.TerminateContainer(dmrCtr); err != nil {
        log.Printf("failed to terminate container: %s", err)
    }
}()
if err != nil {
    log.Printf("failed to start container: %s", err)
    return
}

Module Reference¶

Run function¶

Since v0.37.0

The Docker Model Runner module exposes two entrypoint functions to create the Docker Model Runner container:

Run¶

This function receives two parameters:

func Run(ctx context.Context, opts ...testcontainers.ContainerCustomizer) (*Container, error)

context.Context, the Go context.
testcontainers.ContainerCustomizer, a variadic argument for passing options.

Info

This function will use the default socat image under the hood. Please refer to the socat module for more information.

Image¶

Use the second argument in the Run function to set a valid Docker image. In example: Run(context.Background(), "alpine/socat:1.8.0.1").

Container Options¶

When starting the Docker Model Runner container, you can pass options in a variadic way to configure it.

WithModel¶

Since v0.37.0

Use the WithModel option to set the model to pull when the container is started. Please be aware, that only Models as OCI Artifacts are compatible with Docker Model Runner.

dockermodelrunner.WithModel("ai/llama3.2:latest")

Warning

Multiple calls to this function overrides the previous value.

You can find a curated collection of cutting-edge AI models as OCI Artifacts, from lightweight on-device models to high-performance LLMs on Docker Hub.

The following options are exposed by the testcontainers package.

Basic Options¶

WithExposedPorts Since v0.37.0
WithEnv Since v0.29.0
WithWaitStrategy Since v0.20.0
WithAdditionalWaitStrategy Since v0.38.0
WithWaitStrategyAndDeadline Since v0.20.0
WithAdditionalWaitStrategyAndDeadline Since v0.38.0
WithEntrypoint Since v0.37.0
WithEntrypointArgs Since v0.37.0
WithCmd Since v0.37.0
WithCmdArgs Since v0.37.0
WithLabels Since v0.37.0

Lifecycle Options¶

WithLifecycleHooks Since v0.38.0
WithAdditionalLifecycleHooks Since v0.38.0
WithStartupCommand Since v0.25.0
WithAfterReadyCommand Since v0.28.0

Files & Mounts Options¶

Build Options¶

WithDockerfile Since v0.37.0

Logging Options¶

Image Options¶

Networking Options¶

Advanced Options¶

WithHostPortAccess Since v0.31.0
WithConfigModifier Since v0.20.0
WithHostConfigModifier Since v0.20.0
WithEndpointSettingsModifier Since v0.20.0
CustomizeRequest Since v0.20.0
WithName Since v0.38.0
WithNoStart Since v0.38.0
WithProvider Not available until the next release main

Experimental Options¶

WithReuseByName Since v0.37.0

Container Methods¶

The Docker Model Runner container exposes the following methods:

PullModel¶

Since v0.37.0

Use the PullModel method to pull a model from the Docker Model Runner. Make sure the passed context is not done before the pull operation is completed, so that the pull operation is cancelled.

Pulling a model at runtime

const (
    modelNamespace = "ai"
    modelName      = "smollm2"
    modelTag       = "360M-Q4_K_M"
    fqModelName    = modelNamespace + "/" + modelName + ":" + modelTag
)

ctx, cancel := context.WithTimeout(ctx, 60*time.Second)
defer cancel()

err = dmrCtr.PullModel(ctx, fqModelName)
if err != nil {
    log.Printf("failed to pull model: %s", err)
    return
}

Info

You can find a curated collection of cutting-edge AI models as OCI Artifacts, from lightweight on-device models to high-performance LLMs on Docker Hub.

InspectModel¶

Since v0.37.0

Use the InspectModel method to inspect a model from the Docker Model Runner, by providing the model namespace and name.

Getting a model at runtime

err = dmrCtr.PullModel(ctx, modelNamespace+"/"+modelName+":"+modelTag)
if err != nil {
    log.Printf("failed to pull model: %s", err)
    return
}

model, err := dmrCtr.InspectModel(ctx, modelNamespace, modelName+":"+modelTag)
if err != nil {
    log.Printf("failed to get model: %s", err)
    return
}

The namespace and name of the model is in the format of <name>:<tag>, which defines Models as OCI Artifacts in Docker Hub, therefore the namespace is the organization and the name is the repository.

E.g. ai/smollm2:360M-Q4_K_M. See Models as OCI Artifacts for more information.

ListModels¶

Since v0.37.0

Use the ListModels method to list all models that are already pulled locally, using the Docker Model Runner format.

Listing all models

err = dmrCtr.PullModel(ctx, modelNamespace+"/"+modelName+":"+modelTag)
if err != nil {
    log.Printf("failed to pull model: %s", err)
    return
}

models, err := dmrCtr.ListModels(ctx)
if err != nil {
    log.Printf("failed to get model: %s", err)
    return
}