Register models

The MLflow Model Registry is a central store for versioned models. You can register a trained model from a dev environment, then retrieve any version of it in another dev environment or application. All registered models are accessible to every user on the platform.

Register a model

To register a model, log it as an artifact during a run and then call mlflow.register_model. The registry assigns a version number automatically each time you register under the same model name.

import mlflow
import mlflow.sklearn
from sklearn.linear_model import LogisticRegression

mlflow.set_experiment("shared/my-experiment")

with mlflow.start_run() as run:
    model = LogisticRegression()
    model.fit(X_train, y_train)

    mlflow.sklearn.log_model(model, artifact_path="model")

    run_id = run.info.run_id

model_uri = f"runs:/{run_id}/model"
mlflow.register_model(model_uri=model_uri, name="my-classifier")

The model appears in the Models section of the MLflow UI under the name you provided.

Load a registered model

You can load any registered model version by name. Use this to run inference or continue training from a saved checkpoint.

Load the latest version:

import mlflow.sklearn

model = mlflow.sklearn.load_model("models:/my-classifier/latest")
predictions = model.predict(X_test)

Load a specific version:

model = mlflow.sklearn.load_model("models:/my-classifier/3")

View registered models

Open the MLflow UI from the Developer Workspace sidebar to browse all registered models, compare versions, and review their associated runs.

Next steps

• Open the MLflow UI to browse the Model Registry and compare runs.