Bitzu

Machine Learning Associate ยท MLflow tracking and the model registry

MLflow UI and picking the best run

What the MLflow UI shows (runs table, metrics charts, parameters, artifacts, comparison) and picking the best run programmatically with search_runs and order_by.

Key points

  • The experiment page lists runs in a table with their parameters and metrics, charts metrics over steps, lets you tick runs and compare them, and shows artifacts per run.
  • In code, mlflow.search_runs(experiment_ids=[...], order_by=["metrics.rmse ASC"], max_results=1) returns the best run as a pandas DataFrame (default) or a list of Run objects.

Source: https://mlflow.org/docs/latest/python_api/mlflow.html

Practice questions

Single choice

Which view in the MLflow UI is designed to put several selected runs side by side with their parameters and metrics?

  1. The compare runs view
  2. The artifact browser of one run
  3. The model version page
  4. The cluster event log
Show answer

Answer: A. The compare runs view

The compare view takes runs ticked in the table. The artifact browser shows files of a single run, and the registry lists model versions, not runs.

Select all that apply

Which information can you find for a run in the MLflow UI?

  1. The logged parameters
  2. Metric values and their charts
  3. Every row of the training table
  4. The logged artifacts, such as plots
  5. The invoice cost of the cluster
Show answer

Answer: A. The logged parameters, B. Metric values and their charts, D. The logged artifacts, such as plots

A run page shows parameters, metrics and artifacts. Source data row-level contents and cluster billing are not run details in MLflow.

True or false

Without an order_by argument, mlflow.search_runs returns runs sorted by best metric value.

Show answer

Answer: False

False. The default sort is by start time, newest first, not by any metric. You must pass order_by to rank by a metric.

Practice 8 questions on this topic in Bitzu

Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.

Download on the App StoreGet it on Google Play

More Machine Learning Associate topics in this module