MLflow Model Registry in MLOps - Mini Project: Build & Apply

MLflow Model Registry

📖 Scenario: You are working as a data scientist managing machine learning models. You want to keep track of different versions of your models and their stages like Staging or Production. MLflow Model Registry helps you organize and control your models easily.

🎯 Goal: Build a simple MLflow Model Registry workflow to register a model, add a description, transition its stage, and list all registered models.

📋 What You'll Learn

Use MLflow's Python API

Create a registered model named SampleModel

Add a description to the registered model

Create a model version from a local model path

Transition the model version to Staging stage

List all registered models and their latest versions

💡 Why This Matters

🌍 Real World

MLflow Model Registry is used in real projects to manage machine learning models' lifecycle, track versions, and control deployment stages.

💼 Career

Understanding MLflow Model Registry is important for MLOps engineers and data scientists to maintain model quality and deployment readiness.

1

Create a registered model named SampleModel

Write code to create a registered model called SampleModel using MLflow's client API. Use mlflow.tracking.MlflowClient() to create the client and create_registered_model method to create the model.

MLOps

# Your code here to create MLflow client and registered model

Need a hint?

Use mlflow.tracking.MlflowClient() to create a client object. Then call create_registered_model with the name "SampleModel".

2

Add a description to the registered model SampleModel

Add a description "This is a sample MLflow registered model." to the registered model SampleModel using the update_registered_model method of the MLflow client.

MLOps

import mlflow
client = mlflow.tracking.MlflowClient()
client.create_registered_model(name="SampleModel")
# Your code here to update the model description

Need a hint?

Use update_registered_model with the name and description parameters.

3

Create a model version from a local model path

Create a model version for SampleModel using the local model path "./model" and set the version description to "Initial version". Use the create_model_version method of the MLflow client.

MLOps

import mlflow
client = mlflow.tracking.MlflowClient()
client.create_registered_model(name="SampleModel")
client.update_registered_model(name="SampleModel", description="This is a sample MLflow registered model.")
# Your code here to create a model version

Need a hint?

Use create_model_version with name, source, and description. The run_id can be None if not available.

4

Transition the model version to Staging and list all registered models

Use transition_model_version_stage to move the model version to the Staging stage. Then list all registered models using list_registered_models and print each model's name and latest version number.

MLOps

import mlflow
client = mlflow.tracking.MlflowClient()
client.create_registered_model(name="SampleModel")
client.update_registered_model(name="SampleModel", description="This is a sample MLflow registered model.")
model_version = client.create_model_version(name="SampleModel", source="./model", run_id=None, description="Initial version")
# Your code here to transition stage and list models

Need a hint?

Use transition_model_version_stage with name, version, and stage="Staging". Then call list_registered_models and print the model name and latest version.