Getting started

Installation

Install merlin-sdk from artifactory by running this command:

pip install merlin-sdk

Authenticate to gcloud:

gcloud auth application-default login

Now you are ready to deploy your machine learning model. See sample notebook to get you started.

Concept

Project

Project represent a namespace for a collection of model. In subsequent Merlin release, Project would be the main building block for access control.

Model

Model represent machine learning model. A model can have a type which determine how the model can be deployed. Merlin supports both standard model (XGBoost, SKLearn, Tensorflow, and PyTorch) and user-defined model (PyFunc model). Conceptually, model in Merlin is similar to a class in programming language. To instantiate a model you’ll have to create a model version.

Model Version

Model version represents a snapshot of particular model iteration. A model version might contain artifacts which is deployable to Merlin. Each of a model version will have a version endpoint. Merlin supports up to 3 model version to be deployed at the same time.

Version Endpoint

Version endpoint is URL associated with a model version deployment. Version endpoint URL has following template

http://<model_name>-<version>.<project_name>.<merlin_base_url>

For example a model named mymodel within project named myproject will have a version endpoint for version 1 which look as follow:

http://mymodel-1.myproject.models.id.merlin.dev

Version endpoint has several state:

pending

Is the initial state of a version endpoint.

ready

Once deployed, a version endpoint is in ready state and is accessible.

serving

A version endpoint is in serving state if Model Endpoint has traffic rule which uses the particular Version Endpoint. A model version could not be undeployed if its version endpoint is in serving state.

terminated

Once undeployed a version endpoint is in terminated state.

failed

If error occurred during deployment.

Model Endpoint

Model Endpoint is a stable URL associated with a model. Model endpoint URL has following template

http://<model_name>.<project_name>.<merlin_base_url>

For example a model named mymodel within project named myproject will have model endpoint which look as follow:

http://mymodel.myproject.models.id.merlin.dev

Model endpoint can have a traffic rule which determine which model version will receive traffic when request is received.