How to Install MLflow in Python Step by Step

MLflow is a powerful tool for managing machine learning workflows. It helps track experiments, package code, and deploy models. This guide will walk you through installing MLflow in Python.

Prerequisites

Before installing MLflow, ensure you have Python installed. You can check this by running:

 
python --version

Python 3.8.5

If you don't have Python, download it from the official website. Also, ensure pip is installed. Pip is Python's package installer.

Install MLflow Using Pip

The easiest way to install MLflow is using pip. Open your terminal or command prompt and run:

 
pip install mlflow

Successfully installed mlflow-1.30.0

This command installs the latest version of MLflow. If you encounter a ModuleNotFoundError, check our guide on How To Solve ModuleNotFoundError.

Verify the Installation

After installation, verify MLflow is installed correctly. Run the following command:

 
mlflow --version

mlflow, version 1.30.0

If you see the version number, MLflow is installed successfully.

Start the MLflow Tracking Server

MLflow includes a tracking server to log experiments. Start it by running:

 
mlflow ui

[INFO] Starting gunicorn 20.1.0

This starts the server on http://localhost:5000. Open this URL in your browser to access the MLflow UI.

Basic MLflow Example

Here’s a simple example to test MLflow. Create a Python script named example.py:

 
import mlflow

# Start an MLflow run
with mlflow.start_run():
    # Log a parameter (key-value pair)
    mlflow.log_param("param1", 5)

    # Log a metric
    mlflow.log_metric("metric1", 0.85)

    # Log a tag
    mlflow.set_tag("tag1", "First experiment")

Run the script using:

 
python example.py

Check the MLflow UI at http://localhost:5000. You’ll see your logged experiment.

Install MLflow with Conda

If you use Conda, you can install MLflow in a new environment. Run:

 
conda create -n mlflow-env python=3.8
conda activate mlflow-env
pip install mlflow

This creates a dedicated environment for MLflow. It helps avoid conflicts with other packages.

Conclusion

Installing MLflow in Python is straightforward. Use pip or Conda to set it up. Verify the installation and start the tracking server. Now you’re ready to manage your machine learning workflows efficiently.