MLflow UI

As explained in Machine Learning Platform Engineering

The MLflow UI is a web interface for browsing experiments, runs, parameters, metrics, datasets, and artifacts recorded by MLflow.

MLflow UI provides an operational view of tracked ML work. A local tracking server exposes experiments, runs, parameters, metrics, datasets, and artifacts for inspection and comparison. section 4.1.2

mlflow ui
[2023-09-22 13:50:58 +0800] [16047] [INFO] Starting gunicorn 21.2.0
[2023-09-22 13:50:58 +0800] [16047] [INFO] Listening at: http://127.0.0.1:5000

Start the local interface with mlflow ui, then open the server at http://localhost:5000. section 4.1.2

Use the UI as the visible layer of an MLflow tracking workflow:

• Set the tracking URI and create an experiment such as income-classifier. section 4.1.2
• Inspect each run through its description, datasets, model parameters, metrics, and artifacts. section 4.1.2
• Review plots, model files, datasets, and other disk-backed outputs under the Artifacts tab. section 4.1.2
• Compare candidate models by querying metrics and viewing the results as charts. section 4.1.3
• Open the Models tab to inspect models registered for reuse and reproduction. section 4.1.3

Selecting a candidate model

Query runs with metrics.roc_auc_score_test > 0.8, inspect the chart, and select the model with the highest test AUC. A random forest can then be registered under a named model entry. section 4.1.3

Figure 4.7 Using the MLflow UI to query runs that have a test AUC score > 0.8. The results are displayed in a chart view.

Use the chart view to compare qualifying runs and identify metric differences visually. section 4.1.3

How Machine Learning Platform Engineering covers it

How deep it goes

Shows server startup, experiment navigation, run tabs, artifact inspection, metric queries, and model selection, but not UI internals.

Examples

Uses an income-classifier experiment, exploratory plots, test AUC queries, chart comparisons, and a registered random-forest-classifier model.

What it assumes

Assumes MLflow installation and a tracking setup, while demonstrating experiment creation and run organization from the beginning.

What you can do after

Evaluate runs through the UI, compare metrics, inspect artifacts, and select models for registration or reproduction.

How it connects

Connects MLflow tracking with experiments, runs, metrics, artifacts, datasets, autologging, and the model registry.

liveBookmanning.com

Read Machine Learning Platform Engineering on liveBook Get the book at manning.com

Related concepts

• MLflowpart of · MLflow is an open-source platform for managing the machine-learning lifecycle, including experiment tracking, model and artifact management, and model deployment.
• MLflow clientalternative to · The MLflow client is a programmatic interface for tracking machine-learning runs, managing metadata and artifacts, and accessing registered models.

This explainer was generated by Manning's knowledge graph from the text of Machine Learning Platform Engineering. The section links point to the parts of the book it draws on.