Model focus

As explained in Designing Deep Learning Systems: A guide for software engineers

A metadata-system design approach that organizes and correlates metadata primarily around model files.

Model focus organizes and correlates metadata primarily around model files, helping users trace each model to its training run, inputs, metrics, stages, and artifacts. section 8.3.2

A model-focused metadata system can provide:

• Links from model files to a shared training_run_id, connecting outputs with training data and metrics. section 8.3.2
• Artifact discovery through URLs stored as training-run outputs, including models preserved across test, preproduction, and production stages. section 8.3.2
• A search experience that supports both model-oriented and pipeline-oriented queries, even when storage uses pipeline focus. section 8.3.2

Figure 8.5 An entity relationship diagram of the data schema of the metadata storage

Use the schema diagram to inspect how model files, training runs, inputs, metrics, stages, and artifact URLs are connected. section 8.3.2

How Designing Deep Learning Systems: A guide for software engineers covers it

How deep it goes

Defines model focus and contrasts it with pipeline focus, while deferring detailed implementation to a combined storage and search design.

Examples

Uses model files, training runs, artifact URLs, training metrics, input data, and test-to-production stages as connected entities.

What you can do after

Evaluate metadata architectures by comparing model-centered correlation with pipeline-centered aggregation and supporting both through search.

How it connects

Connects model files with metadata stores, artifact stores, training runs, training data, metrics, and pipeline search.

liveBookmanning.com

Read Designing Deep Learning Systems: A guide for software engineers on liveBook Get the book at manning.com

Related concepts

• pipeline focuscontrasts with · A metadata-system design approach that organizes and correlates metadata primarily around pipelines or training runs.

This explainer was generated by Manning's knowledge graph from the text of Designing Deep Learning Systems: A guide for software engineers. The section links point to the parts of the book it draws on.