The Importance of Tracking

In a production ML system, you will have many different datasets and models. It's crucial to be able to track and compare these artifacts to understand how your system is evolving and to debug any issues that arise.

Vertex AI ML Metadata is a fully managed service that allows you to track and analyze the artifacts and metadata generated by your ML workflows.

---

1. Artifacts, Executions, and Contexts

ML Metadata is based on a simple data model:

• Artifacts: These are the "things" that are produced or consumed by your ML workflow, such as datasets, models, and metrics.
• Executions: These are the steps in your ML workflow, such as data preprocessing, model training, and model evaluation.
• Contexts: These are used to group artifacts and executions together, such as a pipeline run.

---

2. Tracking Artifacts in a Pipeline

When you run a pipeline on Vertex AI Pipelines, the artifacts and metadata are automatically tracked in ML Metadata. This allows you to see the entire lineage of your model, from the raw data to the final deployed model.

You can also use the ML Metadata API to manually create and track your own artifacts.

---

3. Comparing Artifacts

You can use the ML Metadata UI to compare different artifacts. For example, you can:

• Compare two datasets: See the differences in their schemas and statistics.
• Compare two models: See the differences in their hyperparameters and evaluation metrics.
• Compare two pipeline runs: See the differences in their inputs, outputs, and execution times.

---

Knowledge Check