{"id":56,"date":"2025-04-06T14:40:11","date_gmt":"2025-04-06T14:40:11","guid":{"rendered":"https:\/\/aiopsschool.com\/blog\/?p=56"},"modified":"2026-02-17T15:22:42","modified_gmt":"2026-02-17T15:22:42","slug":"mlflow-components","status":"publish","type":"post","link":"https:\/\/aiopsschool.com\/blog\/mlflow-components\/","title":{"rendered":"MLFlow Components"},"content":{"rendered":"\n
\"\"<\/figure>\n\n\n\n

Here’s a short, simple introduction to each of the 4 core components<\/strong> of MLflow:<\/p>\n\n\n\n

\n\n\n\n

\ud83d\ude80 1. MLflow Tracking<\/strong><\/h2>\n\n\n\n
    \n
  • What it does:<\/strong> Logs and tracks your ML experiments.<\/li>\n\n\n\n
  • Benefit:<\/strong> Easily compare different runs, parameters, metrics, and artifacts.<\/li>\n<\/ul>\n\n\n\n
    \n\n\n\n

    \ud83d\udce6 2. MLflow Projects<\/strong><\/h2>\n\n\n\n
      \n
    • What it does:<\/strong> Provides a standardized way to package and run ML code.<\/li>\n\n\n\n
    • Benefit:<\/strong> Ensures reproducibility and simplifies running code across environments.<\/li>\n<\/ul>\n\n\n\n
      \n\n\n\n

      \ud83d\udccc 3. MLflow Models<\/strong><\/h2>\n\n\n\n
        \n
      • What it does:<\/strong> Standardizes packaging and deploying trained models across ML frameworks.<\/li>\n\n\n\n
      • Benefit:<\/strong> Easily deploy, load, and reuse models anywhere.<\/li>\n<\/ul>\n\n\n\n
        \n\n\n\n

        \ud83d\udcda 4. MLflow Model Registry<\/strong><\/h2>\n\n\n\n
          \n
        • What it does:<\/strong> Centralized hub for managing and versioning ML models.<\/li>\n\n\n\n
        • Benefit:<\/strong> Clearly manage model lifecycle stages, simplifying collaboration and deployment.<\/li>\n<\/ul>\n\n\n\n
          \n\n\n\n

          Together, these components simplify managing the entire machine learning lifecycle.<\/p>\n\n\n\n

          MLflow Tracking (Simplified Explanation)<\/strong><\/h3>\n\n\n\n

          MLflow Tracking is a component of MLflow<\/strong> designed specifically to help you keep a clear record of your machine learning experiments.<\/p>\n\n\n\n

          \n\n\n\n

          \ud83d\udd0d What exactly is MLflow Tracking?<\/strong><\/h3>\n\n\n\n

          When you develop ML models, you often:<\/p>\n\n\n\n

            \n
          • Run multiple experiments<\/strong> with different parameters, data, or algorithms.<\/li>\n\n\n\n
          • Need to remember<\/strong> how each experiment performed.<\/li>\n\n\n\n
          • Want an easy way<\/strong> to compare results later.<\/li>\n<\/ul>\n\n\n\n

            MLflow Tracking does exactly that. It automatically records<\/strong> important details about each experiment, including:<\/p>\n\n\n\n

              \n
            • Parameters<\/strong> (e.g., hyperparameters, algorithm choices)<\/li>\n\n\n\n
            • Metrics<\/strong> (e.g., accuracy, precision, recall)<\/li>\n\n\n\n
            • Artifacts<\/strong> (e.g., trained models, plots, reports)<\/li>\n\n\n\n
            • Environment information<\/strong> (e.g., Python version, libraries)<\/li>\n<\/ul>\n\n\n\n
              \n\n\n\n

              \ud83d\udccc How MLflow Tracking works:<\/strong><\/h3>\n\n\n\n

              1. Log Parameters:<\/strong><\/p>\n\n\n\n

                \n
              • Stores hyperparameters or any configuration you used in experiments.<\/li>\n\n\n\n
              • Example:<\/strong>
                mlflow.log_param(\"num_layers\", 4)<\/code><\/li>\n<\/ul>\n\n\n\n

                2. Log Metrics:<\/strong><\/p>\n\n\n\n

                  \n
                • Captures model performance scores and metrics.<\/li>\n\n\n\n
                • Example:<\/strong>
                  mlflow.log_metric(\"accuracy\", 0.95)<\/code><\/li>\n<\/ul>\n\n\n\n

                  3. Log Artifacts:<\/strong><\/p>\n\n\n\n

                    \n
                  • Saves physical files (models, plots, images, datasets, etc.).<\/li>\n\n\n\n
                  • Example:<\/strong>
                    mlflow.log_artifact(\"model.pkl\")<\/code><\/li>\n<\/ul>\n\n\n\n
                    \n\n\n\n

                    \ud83d\udcca MLflow Tracking UI (Dashboard):<\/strong><\/h3>\n\n\n\n

                    MLflow provides an easy-to-use web-based user interface<\/strong>:<\/p>\n\n\n\n

                      \n
                    • Visualize all experiments in a structured way.<\/li>\n\n\n\n
                    • Compare parameters and metrics across runs.<\/li>\n\n\n\n
                    • Download or preview logged artifacts easily.<\/li>\n<\/ul>\n\n\n\n
                      \n\n\n\n

                      \ud83d\ude80 Benefits of using MLflow Tracking:<\/strong><\/h3>\n\n\n\n
                        \n
                      • Easy Experiment Management:<\/strong> Clearly see what worked and what didn’t.<\/li>\n\n\n\n
                      • Improved Collaboration:<\/strong> Share your experiments and results easily with your team.<\/li>\n\n\n\n
                      • Reproducibility:<\/strong> Quickly reproduce any past experiment with exact settings.<\/li>\n\n\n\n
                      • Faster Iteration:<\/strong> Saves time by simplifying experiment comparisons and selection.<\/li>\n<\/ul>\n\n\n\n
                        \n\n\n\n

                        \ud83d\udccc Example scenario:<\/strong><\/h3>\n\n\n\n

                        Imagine you’re training a model to classify emails as spam or not-spam. You try multiple experiments, varying parameters like learning rate, model depth, and epochs.<\/p>\n\n\n\n

                        With MLflow Tracking:<\/p>\n\n\n\n

                          \n
                        • Each experiment\u2019s settings and results are logged automatically.<\/li>\n\n\n\n
                        • You can easily identify your best-performing experiment (highest accuracy).<\/li>\n\n\n\n
                        • Reproduce that exact experiment easily in the future.<\/li>\n<\/ul>\n\n\n\n
                          \n\n\n\n

                          \ud83d\udcdd Simple Python Example using MLflow Tracking:<\/strong><\/h3>\n\n\n\n
                          import mlflow\n\n# Start tracking experiment\nwith mlflow.start_run():\n    mlflow.log_param(\"learning_rate\", 0.01)\n    mlflow.log_param(\"epochs\", 100)\n\n    # Train your model here\n    accuracy = 0.92  # obtained after training\n\n    mlflow.log_metric(\"accuracy\", accuracy)\n\n    # Save your trained model\n    mlflow.log_artifact(\"trained_model.pkl\")\n<\/code><\/pre>\n\n\n\n
                          This simple example clearly illustrates how MLflow Tracking can help you stay organized and focused.<\/p>\n\n\n\n
                          \n\n\n\n
                          \ud83c\udfaf In short:<\/strong><\/h3>\n\n\n\n
                          MLflow Tracking<\/strong> simplifies how you manage, record, visualize, and reproduce your machine learning experiments, helping you build models better and faster.<\/p>\n\n\n\n
                          \ud83d\udccc MLflow Projects (Simplified Explanation)<\/strong><\/h2>\n\n\n\n
                          MLflow Projects<\/strong> are a component of MLflow<\/strong> that help you organize, share, and run your machine learning code consistently across different environments and systems.<\/p>\n\n\n\n
                          \n\n\n\n
                          \ud83c\udfaf What exactly is an MLflow Project?<\/strong><\/h2>\n\n\n\n
                          An MLflow Project<\/strong> is essentially a standardized way of packaging your machine learning code. Think of it as creating a recipe<\/strong> that clearly defines:<\/p>\n\n\n\n
                            \n
                          • Code:<\/strong> The scripts or notebooks that you use to train and evaluate your models.<\/li>\n\n\n\n
                          • Dependencies:<\/strong> Python libraries and other requirements to run your code.<\/li>\n\n\n\n
                          • Execution Environment:<\/strong> Details on how and where your code should be run (e.g., locally, cloud, containers).<\/li>\n\n\n\n
                          • Parameters:<\/strong> Input values that can be changed when running your code (like learning rate, dataset location).<\/li>\n<\/ul>\n\n\n\n
                            \n\n\n\n
                            \ud83d\udd27 Why use MLflow Projects?<\/strong><\/h2>\n\n\n\n
                            Machine learning code often runs into problems like:<\/p>\n\n\n\n
                              \n
                            • “It works on my machine”<\/strong> (but not elsewhere).<\/li>\n\n\n\n
                            • Dependency issues<\/strong> (Python versions, library conflicts).<\/li>\n\n\n\n
                            • Reproducibility issues<\/strong> (difficulty repeating results exactly).<\/li>\n<\/ul>\n\n\n\n
                              MLflow Projects solve these issues by providing a standardized format and easy execution method.<\/p>\n\n\n\n
                              \n\n\n\n
                              \ud83d\udce6 Structure of an MLflow Project<\/strong><\/h2>\n\n\n\n
                              A typical MLflow project contains:<\/p>\n\n\n\n
                              my-ml-project\/\n\u251c\u2500\u2500 MLproject           # Specifies project configuration\n\u251c\u2500\u2500 conda.yaml          # Defines dependencies (optional)\n\u251c\u2500\u2500 train.py            # Your training script\n\u251c\u2500\u2500 data\/               # Data files (optional)\n\u2514\u2500\u2500 README.md           # Documentation (optional)\n<\/code><\/pre>\n\n\n\n
                              \n\n\n\n
                              \ud83d\udcc4 Example: MLproject file<\/strong><\/h2>\n\n\n\n
                              The MLproject<\/code> file is the core. It describes how to run your project:<\/p>\n\n\n\n
                              name: SpamClassifier\n\nconda_env: conda.yaml\n\nentry_points:\n  main:\n    command: \"python train.py --learning_rate {learning_rate} --epochs {epochs}\"\n    parameters:\n      learning_rate: {type: float, default: 0.01}\n      epochs: {type: int, default: 100}\n<\/code><\/pre>\n\n\n\n
                                \n
                              • name<\/code>: Identifies your project.<\/li>\n\n\n\n
                              • conda_env<\/code>: Points to your environment dependencies file.<\/li>\n\n\n\n
                              • entry_points<\/code>: Specifies the command and parameters needed to run your code.<\/li>\n<\/ul>\n\n\n\n
                                \n\n\n\n
                                \ud83d\ude80 Running MLflow Projects<\/strong><\/h2>\n\n\n\n
                                Run a project easily using:<\/p>\n\n\n\n
                                mlflow run .\/my-ml-project -P learning_rate=0.005 -P epochs=50\n<\/code><\/pre>\n\n\n\n
                                MLflow automatically:<\/p>\n\n\n\n
                                  \n
                                • Sets up the correct environment (installs libraries from conda.yaml<\/code>).<\/li>\n\n\n\n
                                • Runs your code with specified parameters.<\/li>\n<\/ul>\n\n\n\n
                                  \n\n\n\n
                                  \ud83c\udf81 Benefits of MLflow Projects<\/strong><\/h2>\n\n\n\n
                                    \n
                                  • Reproducibility:<\/strong> Easily recreate the exact conditions to run experiments.<\/li>\n\n\n\n
                                  • Portability:<\/strong> Run your ML code consistently across local machines, cloud platforms, or containers.<\/li>\n\n\n\n
                                  • Collaboration:<\/strong> Share your projects seamlessly with your team or community.<\/li>\n<\/ul>\n\n\n\n
                                    \n\n\n\n
                                    \ud83c\udf10 Common Usage Examples<\/strong><\/h2>\n\n\n\n
                                      \n
                                    • Packaging your ML experiments clearly.<\/li>\n\n\n\n
                                    • Running experiments repeatedly on different datasets.<\/li>\n\n\n\n
                                    • Deploying experiments easily to cloud environments (AWS, Azure, GCP).<\/li>\n\n\n\n
                                    • Creating consistent workflows for ML pipelines.<\/li>\n<\/ul>\n\n\n\n
                                      \n\n\n\n
                                      \ud83d\udccc In short:<\/strong><\/h2>\n\n\n\n
                                      MLflow Projects<\/strong> help you standardize, organize, and run your machine learning code consistently, solving reproducibility, dependency, and collaboration challenges commonly encountered in ML workflows.<\/p>\n\n\n\n
                                      \ud83d\udccc MLflow Models (Simplified Explanation)<\/strong><\/h2>\n\n\n\n
                                      MLflow Models<\/strong> is a key component in MLflow that helps you package, manage, deploy, and reuse your trained Machine Learning models in a standardized, platform-independent way.<\/p>\n\n\n\n
                                      \n\n\n\n
                                      \ud83c\udfaf What exactly is an MLflow Model?<\/strong><\/h2>\n\n\n\n
                                      An MLflow Model<\/strong> is a standardized format<\/strong> for packaging trained ML models, making it easy to:<\/p>\n\n\n\n
                                        \n
                                      • Save<\/strong> models created with different ML frameworks (TensorFlow, PyTorch, scikit-learn, XGBoost, etc.).<\/li>\n\n\n\n
                                      • Load and reuse<\/strong> models later without confusion.<\/li>\n\n\n\n
                                      • Deploy<\/strong> models to different environments (e.g., REST APIs, batch predictions, cloud platforms).<\/li>\n<\/ul>\n\n\n\n
                                        Think of MLflow Models as a universal container<\/strong> that clearly defines how your trained model can be loaded and used consistently.<\/p>\n\n\n\n
                                        \n\n\n\n
                                        \ud83d\udce6 Structure of MLflow Models<\/strong><\/h2>\n\n\n\n
                                        An MLflow model typically looks like this:<\/p>\n\n\n\n
                                        mlruns\/\n\u2514\u2500\u2500 0\/\n    \u2514\u2500\u2500 run_id\/\n        \u2514\u2500\u2500 artifacts\/\n            \u2514\u2500\u2500 model\/\n                \u251c\u2500\u2500 MLmodel            # MLflow metadata about the model\n                \u251c\u2500\u2500 conda.yaml         # Dependencies required by the model\n                \u2514\u2500\u2500 model.pkl          # The actual serialized model file (example: sklearn)\n<\/code><\/pre>\n\n\n\n
                                          \n
                                        • MLmodel<\/code>: Metadata file<\/strong> describing the model type and how to load it.<\/li>\n\n\n\n
                                        • conda.yaml<\/code>: Specifies environment dependencies required by the model.<\/li>\n\n\n\n
                                        • model.pkl<\/code>: The actual serialized model file itself.<\/li>\n<\/ul>\n\n\n\n
                                          \n\n\n\n
                                          \ud83d\udd16 The MLmodel file (Important)<\/strong><\/h2>\n\n\n\n
                                          This file provides metadata about the MLflow model:<\/p>\n\n\n\n
                                          artifact_path: model\nflavors:\n  sklearn:\n    pickled_model: model.pkl\n  python_function:\n    loader_module: mlflow.sklearn\n    python_version: 3.10.6\nrun_id: abc1234\n<\/code><\/pre>\n\n\n\n
                                            \n
                                          • flavors<\/code>: Defines how MLflow can load your model. Each framework is considered a \u201cflavor\u201d (like sklearn, tensorflow, pytorch, etc.).<\/li>\n\n\n\n
                                          • MLflow uses this metadata to load models uniformly across different ML libraries.<\/li>\n<\/ul>\n\n\n\n
                                            \n\n\n\n
                                            \ud83d\ude80 MLflow Model Flavors<\/strong><\/h2>\n\n\n\n
                                            MLflow supports a variety of ML libraries as flavors<\/strong>, including:<\/p>\n\n\n\n
                                              \n
                                            • Scikit-learn<\/strong><\/li>\n\n\n\n
                                            • TensorFlow & Keras<\/strong><\/li>\n\n\n\n
                                            • PyTorch<\/strong><\/li>\n\n\n\n
                                            • XGBoost<\/strong><\/li>\n\n\n\n
                                            • LightGBM<\/strong><\/li>\n\n\n\n
                                            • CatBoost<\/strong><\/li>\n\n\n\n
                                            • Spark MLlib<\/strong><\/li>\n\n\n\n
                                            • Prophet<\/strong><\/li>\n\n\n\n
                                            • ONNX<\/strong><\/li>\n<\/ul>\n\n\n\n
                                              Each flavor allows you to load models consistently<\/strong>, even across different frameworks.<\/p>\n\n\n\n
                                              \n\n\n\n
                                              \ud83d\udccc How MLflow Models Help<\/strong><\/h2>\n\n\n\n
                                                \n
                                              • Unified Serving<\/strong>: Deploy your model using REST APIs or batch prediction pipelines seamlessly.<\/li>\n\n\n\n
                                              • Interoperability<\/strong>: Easily switch between ML frameworks, simplifying your workflow.<\/li>\n\n\n\n
                                              • Easy Deployment<\/strong>: Deploy models to cloud services (AWS SageMaker, Azure ML, Google Cloud Vertex AI) without extra hassle.<\/li>\n\n\n\n
                                              • Reproducibility<\/strong>: Clearly documents model dependencies and environment for reproducibility.<\/li>\n<\/ul>\n\n\n\n
                                                \n\n\n\n
                                                \ud83e\uddd1\u200d\ud83d\udcbb Simple Example: Save & Load MLflow Model (with sklearn)<\/strong><\/h2>\n\n\n\n
                                                import mlflow.sklearn\nfrom sklearn.ensemble import RandomForestClassifier\n\n# Train model\nmodel = RandomForestClassifier()\nmodel.fit(X_train, y_train)\n\n# Save model using MLflow\nmlflow.sklearn.save_model(model, \"rf_model\")\n\n# Load model again later\nloaded_model = mlflow.sklearn.load_model(\"rf_model\")\n\n# Make predictions easily\npredictions = loaded_model.predict(X_test)\n<\/code><\/pre>\n\n\n\n
                                                \n\n\n\n
                                                \ud83c\udf10 Deploying MLflow Models<\/strong><\/h2>\n\n\n\n
                                                MLflow Models can be easily deployed as REST APIs:<\/p>\n\n\n\n
                                                mlflow models serve -m rf_model -p 1234\n<\/code><\/pre>\n\n\n\n
                                                Now, your model is accessible at:<\/p>\n\n\n\n
                                                http://localhost:1234\/invocations\n<\/code><\/pre>\n\n\n\n
                                                \n\n\n\n
                                                \ud83c\udf81 Benefits Summary<\/strong><\/h2>\n\n\n\n
                                                  \n
                                                • Easy Packaging<\/strong>: Clearly packages models across different ML frameworks.<\/li>\n\n\n\n
                                                • Reusability<\/strong>: Easy reuse and deployment of trained models.<\/li>\n\n\n\n
                                                • Portability<\/strong>: Consistent loading and serving across diverse environments.<\/li>\n\n\n\n
                                                • Collaboration<\/strong>: Simplifies sharing models with teammates and deployment engineers.<\/li>\n<\/ul>\n\n\n\n
                                                  \n\n\n\n
                                                  \ud83d\udccc In short:<\/strong><\/h2>\n\n\n\n
                                                  MLflow Models<\/strong> make it easy to package, reuse, and deploy trained machine learning models consistently, solving common ML challenges related to deployment, portability, and reuse.<\/p>\n\n\n\n
                                                  MLflow Models and MLflow Projects<\/h2>\n\n\n\n
                                                  Let’s simplify clearly:<\/p>\n\n\n\n
                                                  \ud83d\udd0d Simple difference:<\/strong><\/h2>\n\n\n\n
                                                    \n
                                                  • MLflow Projects:<\/strong>\n
                                                      \n
                                                    • Standardize and package your ML code<\/strong> for running training, testing, and experiments.<\/li>\n\n\n\n
                                                    • Think of it as a “recipe<\/strong>” for executing ML code consistently.<\/li>\n<\/ul>\n<\/li>\n\n\n\n
                                                    • MLflow Models:<\/strong>\n
                                                        \n
                                                      • Standardize and package your trained ML models<\/strong> to make deploying and reusing easy.<\/li>\n\n\n\n
                                                      • Think of it as “packaging your finished product<\/strong>” (trained model) for easy use.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n\n\n\n
                                                        \n\n\n\n
                                                        \ud83d\uddc2\ufe0f Detailed Comparison:<\/strong><\/h2>\n\n\n\n
                                                        Feature<\/th>MLflow Projects<\/strong><\/th>MLflow Models<\/strong><\/th><\/tr><\/thead>
                                                        What it manages<\/strong><\/td>Code<\/strong> (scripts, dependencies, parameters)<\/td>Trained Models<\/strong> (TensorFlow, PyTorch, Scikit-learn, etc.)<\/td><\/tr>
                                                        Main purpose<\/strong><\/td>Ensure reproducibility of experiments<\/strong><\/td>Ensure consistent deployment and reuse<\/strong> of trained models<\/td><\/tr>
                                                        Output<\/strong><\/td>Results (metrics, artifacts, trained models)<\/td>Packaged trained model ready for deployment<\/td><\/tr>
                                                        Analogy<\/strong><\/td>Recipe (Instructions to cook food)<\/td>Packaged meal (Ready-to-eat food)<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
                                                        \n\n\n\n
                                                        \ud83c\udfaf Real-life Example:<\/strong><\/h2>\n\n\n\n
                                                          \n
                                                        • MLflow Project:<\/strong>
                                                          You define a standardized set of instructions (scripts, dependencies) to run your ML training.<\/li>\n\n\n\n
                                                        • MLflow Model:<\/strong>
                                                          After training your model using the instructions above, you save and package the trained model to be easily deployed as an API.<\/li>\n<\/ul>\n\n\n\n
                                                          \n\n\n\n
                                                          \ud83d\ude80 In short:<\/strong><\/h2>\n\n\n\n
                                                            \n
                                                          • MLflow Project<\/strong> \u2192 Standardize ML code execution<\/strong> (running experiments).<\/li>\n\n\n\n
                                                          • MLflow Model<\/strong> \u2192 Standardize trained model packaging<\/strong> (deploying models).<\/li>\n<\/ul>\n\n\n\n
                                                            This clear distinction makes MLflow powerful and organized in managing the full lifecycle of ML projects.<\/p>\n\n\n\n
                                                            \ud83d\udccc MLflow Model Registry (Simplified Explanation)<\/strong><\/h2>\n\n\n\n
                                                            The MLflow Model Registry<\/strong> is a centralized place to store, organize, version, and manage<\/strong> your machine learning models within MLflow. It simplifies how you keep track of different models, their lifecycle stages, and how you deploy them.<\/p>\n\n\n\n
                                                            \n\n\n\n
                                                            \ud83c\udfaf What Exactly is MLflow Model Registry?<\/strong><\/h2>\n\n\n\n
                                                            Think of MLflow Model Registry like a model management hub<\/strong>, which clearly helps you:<\/p>\n\n\n\n
                                                              \n
                                                            • Keep track of all models<\/strong> created during experiments.<\/li>\n\n\n\n
                                                            • Version<\/strong> your models (similar to version control systems like Git).<\/li>\n\n\n\n
                                                            • Define and manage lifecycle stages<\/strong> for models (e.g., staging, production, archived).<\/li>\n\n\n\n
                                                            • Facilitate collaboration by allowing teams to easily find and reuse the best models.<\/li>\n<\/ul>\n\n\n\n
                                                              \n\n\n\n
                                                              \ud83d\udce6 Main Components of MLflow Model Registry<\/strong><\/h2>\n\n\n\n
                                                              The MLflow Model Registry provides the following key elements:<\/p>\n\n\n\n
                                                              Component<\/th>Explanation<\/th><\/tr><\/thead>
                                                              Model Name<\/strong><\/td>Identifies your registered model uniquely<\/td><\/tr>
                                                              Model Version<\/strong><\/td>Automatic numbering (e.g., Version 1, 2, 3…)<\/td><\/tr>
                                                              Model Stages<\/strong><\/td>Assign stages such as Staging<\/code>, Production<\/code>, Archived<\/code><\/td><\/tr>
                                                              Annotations<\/strong><\/td>Comments and descriptions for clarity<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
                                                              \n\n\n\n
                                                              \ud83d\udccc Typical Workflow with MLflow Model Registry<\/strong><\/h2>\n\n\n\n
                                                              Here’s how you’d typically use MLflow Model Registry in practice:<\/p>\n\n\n\n
                                                              Step 1: Register the Model<\/strong><\/p>\n\n\n\n
                                                                \n
                                                              • After training and evaluating, register the model in MLflow.<\/li>\n<\/ul>\n\n\n\n
                                                                import mlflow\n\nmlflow.register_model(\n    model_uri=\"runs:\/<run-id>\/model\",\n    name=\"CustomerChurnClassifier\"\n)\n<\/code><\/pre>\n\n\n\n
                                                                Step 2: Versioning Models<\/strong><\/p>\n\n\n\n
                                                                  \n
                                                                • Each time you register, MLflow automatically increments the model version.\n
                                                                    \n
                                                                  • Version 1, Version 2, etc.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n\n\n\n
                                                                    Step 3: Transition Model Stages<\/strong><\/p>\n\n\n\n
                                                                      \n
                                                                    • Move models between lifecycle stages (Staging \u2192 Production \u2192 Archived).<\/li>\n<\/ul>\n\n\n\n
                                                                      client = mlflow.tracking.MlflowClient()\n\n# Promote model to production\nclient.transition_model_version_stage(\n    name=\"CustomerChurnClassifier\",\n    version=3,\n    stage=\"Production\"\n)\n<\/code><\/pre>\n\n\n\n
                                                                      Step 4: Deploy and Serve the Model<\/strong><\/p>\n\n\n\n
                                                                        \n
                                                                      • Easily deploy the registered model to serve predictions.<\/li>\n<\/ul>\n\n\n\n
                                                                        \n\n\n\n
                                                                        \ud83d\udea6 Lifecycle Stages in Model Registry<\/strong><\/h2>\n\n\n\n
                                                                        MLflow provides clear lifecycle management with predefined stages:<\/p>\n\n\n\n
                                                                          \n
                                                                        • \u2705 None<\/strong> (default initial stage after registration)<\/li>\n\n\n\n
                                                                        • \u2705 Staging<\/strong> (model under validation\/testing)<\/li>\n\n\n\n
                                                                        • \u2705 Production<\/strong> (model actively deployed for real-world use)<\/li>\n\n\n\n
                                                                        • \u2705 Archived<\/strong> (older models no longer in active use)<\/li>\n<\/ul>\n\n\n\n
                                                                          This makes it easy to clearly see the status of your models at any time.<\/p>\n\n\n\n
                                                                          \n\n\n\n
                                                                          \ud83d\ude80 Benefits of MLflow Model Registry<\/strong><\/h2>\n\n\n\n
                                                                            \n
                                                                          • Centralized Management<\/strong>: One clear location for all your ML models.<\/li>\n\n\n\n
                                                                          • Simplified Versioning<\/strong>: Tracks all changes and versions automatically.<\/li>\n\n\n\n
                                                                          • Clear Deployment Stages<\/strong>: Clarifies the state of your models (e.g., production-ready, staging, deprecated).<\/li>\n\n\n\n
                                                                          • Improved Collaboration<\/strong>: Makes sharing and reusing models within a team straightforward.<\/li>\n\n\n\n
                                                                          • Easy Model Governance<\/strong>: Enables you to audit, track, and manage models efficiently, supporting compliance needs.<\/li>\n<\/ul>\n\n\n\n
                                                                            \n\n\n\n
                                                                            \ud83c\udf10 Example Scenario<\/strong><\/h2>\n\n\n\n
                                                                            Imagine you’re predicting customer churn. After training various models:<\/p>\n\n\n\n
                                                                              \n
                                                                            • You register your best-performing model as CustomerChurnClassifier<\/code>.<\/li>\n\n\n\n
                                                                            • MLflow automatically assigns Version 1<\/strong>.<\/li>\n\n\n\n
                                                                            • You test Version 1 in the Staging<\/strong> environment.<\/li>\n\n\n\n
                                                                            • When you’re confident, promote Version 1 to Production<\/strong>.<\/li>\n\n\n\n
                                                                            • Later, when an improved model (Version 2) is created, Version 1 can be moved to Archived<\/strong>, and Version 2 promoted to Production<\/strong>.<\/li>\n<\/ul>\n\n\n\n
                                                                              \n\n\n\n
                                                                              \ud83d\udccc MLflow Model Registry UI<\/strong><\/h2>\n\n\n\n
                                                                              MLflow provides a user-friendly web interface<\/strong>:<\/p>\n\n\n\n
                                                                                \n
                                                                              • Clearly view models, versions, and stages.<\/li>\n\n\n\n
                                                                              • Easily transition models between stages.<\/li>\n\n\n\n
                                                                              • Add descriptions, documentation, and notes directly.<\/li>\n<\/ul>\n\n\n\n
                                                                                \n\n\n\n
                                                                                \ud83d\udcdd Quick Python Example<\/strong><\/h2>\n\n\n\n
                                                                                import mlflow\nfrom mlflow.tracking import MlflowClient\n\n# Create MLflow client\nclient = MlflowClient()\n\n# Register the model after training\nmodel_uri = \"runs:\/<run-id>\/model\"\nmodel_name = \"CustomerChurnClassifier\"\nclient.create_registered_model(model_name)\n\nmodel_version = client.create_model_version(\n    name=model_name,\n    source=model_uri,\n    run_id=\"<run-id>\"\n)\n\n# Transition to Production\nclient.transition_model_version_stage(\n    name=model_name,\n    version=model_version.version,\n    stage=\"Production\"\n)\n<\/code><\/pre>\n\n\n\n
                                                                                \n\n\n\n
                                                                                \ud83d\udccc In Short:<\/strong><\/h2>\n\n\n\n
                                                                                The MLflow Model Registry<\/strong> provides a simple, clear, and efficient way to organize, manage, version, and deploy your ML models, making the entire ML lifecycle significantly easier to manage and collaborate on.<\/p>\n\n\n\n
                                                                                <\/p>\n","protected":false},"excerpt":{"rendered":"
                                                                                Here’s a short, simple introduction to each of the 4 core components of MLflow: \ud83d\ude80 1. MLflow Tracking \ud83d\udce6 2. […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[242],"tags":[],"class_list":["post-56","post","type-post","status-publish","format-standard","hentry","category-training"],"_links":{"self":[{"href":"https:\/\/aiopsschool.com\/blog\/wp-json\/wp\/v2\/posts\/56","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/aiopsschool.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/aiopsschool.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/aiopsschool.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/aiopsschool.com\/blog\/wp-json\/wp\/v2\/comments?post=56"}],"version-history":[{"count":3,"href":"https:\/\/aiopsschool.com\/blog\/wp-json\/wp\/v2\/posts\/56\/revisions"}],"predecessor-version":[{"id":60,"href":"https:\/\/aiopsschool.com\/blog\/wp-json\/wp\/v2\/posts\/56\/revisions\/60"}],"wp:attachment":[{"href":"https:\/\/aiopsschool.com\/blog\/wp-json\/wp\/v2\/media?parent=56"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/aiopsschool.com\/blog\/wp-json\/wp\/v2\/categories?post=56"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/aiopsschool.com\/blog\/wp-json\/wp\/v2\/tags?post=56"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}