ML lifecycle stages in MLOps - Commands & Configuration

Introduction

Machine learning projects go through several steps from understanding the problem to deploying a working model. These steps help organize work and make sure the model solves the right problem and works well in real life.

When you want to build a model to predict customer behavior based on past data

When you need to clean and prepare data before training a model

When you want to test different models to find the best one

When you want to deploy a model so it can make predictions in a live app

When you need to monitor a model’s performance after deployment to keep it accurate

Commands

This command runs an ML project using MLflow, which helps track experiments and organize the ML lifecycle steps.

Terminal

mlflow run .

Expected Output

2024/06/01 12:00:00 INFO mlflow.projects: === Run (ID='123abc') succeeded ===

→

--experiment-name - Specify the experiment to track this run under

Starts the MLflow tracking UI so you can see your experiments, runs, and metrics in a web browser.

Terminal

mlflow ui

Expected Output

2024/06/01 12:00:05 INFO mlflow.server: Starting MLflow UI at http://127.0.0.1:5000

→

--port - Change the port where the UI runs

Deploys the trained model from a specific run so it can serve predictions via a REST API on port 1234.

Terminal

mlflow models serve -m runs:/123abc/model -p 1234

Expected Output

2024/06/01 12:00:10 INFO mlflow.models: Serving model at http://127.0.0.1:1234

→

-m - Specify the model path to serve

→

-p - Set the port for the model server

Key Concept

If you remember nothing else, remember: the ML lifecycle organizes work into clear steps from data to deployment to keep projects on track and models reliable.

Common Mistakes

Skipping data preparation before training

Poor data quality leads to bad model performance and unreliable predictions

Always clean and prepare your data carefully before training

Not tracking experiments and parameters

You lose track of what settings produced which results, making it hard to improve

Use tools like MLflow to log parameters, metrics, and models

Deploying models without testing

Untested models may fail or give wrong predictions in real use

Test models thoroughly before deployment and monitor them after

Summary

Run ML projects and track experiments with mlflow run and mlflow ui commands.

Deploy trained models as REST APIs using mlflow models serve.

Organize ML work into stages: data prep, training, testing, deployment, and monitoring.