liveBook · Manning

Metrics artifact type

As explained in Machine Learning Platform Engineering

A metrics artifact type is a KFP artifact type that stores numerical measurements as key-value data.

A Metrics artifact stores numerical measurements as simple key-value data inside a KFP pipeline. Use it for model performance, training, evaluation, and other numerical results. section 7.2.3

A Metrics output fits naturally into evaluation and reporting workflows:

  • Use ClassificationMetrics instead when the evaluation needs classification-specific visualizations such as confusion matrices or ROC curves. section 7.2.3

YOLO validation metrics

A KFP validation component accepts a configuration Artifact, trained Model, and validation Dataset, then writes COCO-style measurements to Output[Metrics]. Ultralytics validation logs map50-95, map50, and map75 through metrics.log_metric(). section 8.1.4

How Machine Learning Platform Engineering covers it

How deep it goes
Defines key-value metrics and demonstrates logging, but defers metric semantics and broader evaluation methodology.
Examples
Uses accuracy, F1 score, COCO mAP values, model reports, and model cards as concrete metrics artifacts.
What it assumes
Assumes familiarity with KFP components, Model and Dataset artifacts, and an evaluation function or validation library.
What you can do after
Apply it via KFP by declaring Output[Metrics] and logging numerical results with metrics.log_metric().
How it connects
Connects Model artifacts, Dataset artifacts, pipeline components, model validation, reporting, and ClassificationMetrics.
Cover of Machine Learning Platform Engineering
liveBookmanning.com
Read Machine Learning Platform Engineering on liveBook Get the book at manning.com

Related concepts

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.