Artifact store

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

A storage system for versioned files and objects produced or consumed by machine learning workflows.

An artifact store retains datasets, model files, and model versions produced by training experiments so they remain persistent and retrievable for serving and analysis. section 8.0 section 6.1.3

A model-serving workflow assigns the store and its neighboring components distinct responsibilities:

• The artifact store holds models produced by model training. section 6.1.3
• The prediction web API receives prediction requests. section 6.1.3
• The prediction service loads a model from the store into memory or a GPU, executes it, and returns the result. section 6.1.3

Beyond serving, persistent artifact and metadata retrieval supports model reproduction, experiment comparison, model selection, and investigation of model degradation. section 8.0

Figure 6.3 A typical model prediction workflow in a prediction service

Use the workflow diagram to trace requests from the prediction API through model loading and execution. section 6.1.3

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

How deep it goes

Explains persistence, retrieval, model loading, component responsibilities, and store design principles, while introducing ML Metadata and MLflow.

Examples

Uses datasets, model files, model versions, training metrics, and a shark-detection prediction service.

What you can do after

Diagnose/debug with it by reproducing models, comparing experiments, selecting candidates, and investigating model degradation.

How it connects

Connects model training, model serving, metadata management, experiment comparison, and troubleshooting.

liveBookmanning.com

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

Related concepts

• Model Servingused by · A model-serving platform is software infrastructure that hosts trained models and exposes inference by accepting requests and returning model outputs.
• artifact repositorykind of · An artifact repository is a service that stores, versions, and distributes software packages, build outputs, container images, or other versioned artifacts.

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.