Customer-managed MLflow Servers (Preview)

You can choose to use your own MLflow tracking servers in your Oracle AI Data Platform Workbench notebooks.

Note:

Built-in integrations like model catalogue, experiments do not work when you use your own MLflow tracking server.

You can choose to integrate your own MLflow tracking server with your notebooks by including this code at the beginning of your notebook:

os.environ.pop("MLFLOW_TRACKING_AUTH", None) 
os.environ["MLFLOW_TRACKING_URI"] = "http://<your-mlflow-host>:5000" 

This code disables the MLFLOW_TRACKING_AUTH variable and sets the tracking URI to your tracking server <your-mlflow-host> prior to running any MLflow commands in the notebook.