Skip to main content
You can export models, experiments and training runs from your Managed MLflow cluster to a directory on your machine:
  1. Configure connection to your cluster by setting the MLFLOW_TRACKING_URI, MLFLOW_TRACKING_USERNAME and MLFLOW_TRACKING_PASSWORD environment variables used by MLflow Tracking.
    Clusters with disabled public access do not have public tracking URIs. You can save the private tracking URI to MLFLOW_TRACKING_URI and use it to export data to a Compute virtual machine that belongs to the same network as the cluster.
  2. Install the MLflow Export Import package: 
    pip install git+https://github.com/mlflow/mlflow-export-import/#egg=mlflow-export-import
  3. Run export-all to export all data from the cluster: 
    export-all --output-dir </output/directory> --use-threads True
    You can also use other tools from the package to export single and multiple objects, or to move them between clusters. For details, see the tools overview and the guide to tools for copying in the package’s documentation.