Manage MLflow

This guide describes how to manage MLflow within your Data Science Stack (DSS) environment.

MLflow is a platform for managing the end-to-end machine learning life cycle. It includes tracking experiments, packaging code into reproducible runs, and sharing and deploying models.

Access MLflow

You can access the MLflow User Interface (UI) within your DSS environment through a web browser, by navigating to the URL associated with MLflow. This UI allows you to interact directly with your MLflow experiments and models.

  1. Get the MLflow URL:

To find the URL of MLflow, run:

dss status

Look for the MLflow URL in the output. For example:

MLflow deployment: Ready
MLflow URL: http://10.152.183.205:5000

Note

To access the UI, your MLflow deployment should be Ready.

  1. Access the MLflow UI:

Once you know the URL, open a web browser and enter the URL into the address bar. This will direct you to the MLflow interface.

Get logs

You can retrieve MLflow logs within your DSS environment. Retrieving logs is a critical task for maintaining and troubleshooting MLflow.

To get MLflow logs, use the dss logs command with the --mlflow option:

dss logs --mlflow

You should expect an output like this:

Logs for mlflow-6bbfc5db5-xlfvj:
[2024-04-30 07:57:54 +0000] [22] [INFO] Starting gunicorn 20.1.0
[2024-04-30 07:57:54 +0000] [22] [INFO] Listening at: http://0.0.0.0:5000 (22)
[2024-04-30 07:57:54 +0000] [22] [INFO] Using worker: sync
[2024-04-30 07:57:54 +0000] [23] [INFO] Booting worker with pid: 23
[2024-04-30 07:57:54 +0000] [24] [INFO] Booting worker with pid: 24
[2024-04-30 07:57:54 +0000] [25] [INFO] Booting worker with pid: 25
[2024-04-30 07:57:54 +0000] [26] [INFO] Booting worker with pid: 26

Get artefacts

MLflow artefacts, including models, experiments and runs are stored within your DSS environment. You can access and download them from the MLflow UI.

See also