Databricks

MLflow Commands

Track experiments, manage models, and deploy ML workflows with MLflow on Databricks. Log metrics, parameters, and artifacts for reproducible machine learning.

6 commands

Pro Tips

Use 'mlflow run' to execute reproducible ML projects with dependency management.

Use MLflow autologging to automatically capture metrics without manual logging code.

Common Mistakes

Large artifacts (models, data) can quickly consume storage. Clean up old experiment runs regularly.

Transitioning model stages affects all users. Coordinate stage transitions for production models.

Commands

List experiments

$ databricks experiments list

List all MLflow experiments.

Create experiment

$ databricks experiments create --name /Users/user@email.com/my-experiment

Create a new MLflow experiment.

Search runs

$ databricks runs search --experiment-ids <exp-id>

Search for runs in an experiment.

List registered models

$ databricks registered-models list

List all registered models in the registry.

Get model details

$ databricks registered-models get --name <model-name>

Get details of a registered model.

Transition model stage

$ databricks model-versions transition-stage --name <model> --version 1 --stage Production

Transition a model version to a new stage.