A no-nonsense demo of MLflow on Databricks
- In Databricks, you can either:
- Download and import this file as a
.py
file - If using Databricks Repos, click "Create Repo" and pass in the public URL for this Github Repo
- Download and import this file as a
- Create a Cluster with (at least) the following settings:
- Run the cells!
- You will see runs for each model appear in the sidebar on the right.
- Click one of these to explore the MLflow UI
- From the Artifacts of a run, click Register Model under a new name (e.g.
mlflow101_model
) - In Model Registry, promote this model to Production, then run the last cell in the notebook using the name of that model (e.g.
mlflow101_model
)