MLflow

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

MLflow is an open-source platform for managing the machine learning lifecycle, including experiment tracking, artifact management, model registration, and deployment.

MLflow is an open-source MLOps platform for managing machine-learning experimentation, reproducibility, deployment, and model registration. It preserves training metadata and artifacts so teams can compare experiments, reproduce models, and troubleshoot performance degradation. section 8.0 section 8.4.2

Organize the platform around four components:

• MLflow Tracking records and queries run metadata such as parameters, metrics, and tags. section 8.4.2
• MLflow Projects packages code for reusable and reproducible execution. section 8.4.2
• MLflow Models packages models for use with different serving tools. section 8.4.2
• MLflow Model Registry manages lineage, versions, annotations, and production promotion through a user interface. section 8.4.2

A common deployment uses a remote tracking server with proxied artifact storage. Pipeline steps send metadata and artifacts through the MLflow client; the server stores metadata in SQL and artifacts such as models, images, and documents in object storage. Proxying avoids giving users direct credentials to the remote store. section 8.4.2

Track a training run

Set the remote tracking URI, select an experiment, start a run, and log parameters, metrics, and files. The client API can also retrieve run metadata, experiments, and model versions through MlflowClient.

Figure 8.7 Setting up MLflow tracker server for metadata ingestion and query

Use the architecture diagram to trace metadata ingestion and artifact access through the tracking server. section 8.4.2

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

How deep it goes

Explains component responsibilities, remote-server architecture, storage boundaries, artifact-upload choices, and client operations, without implementation internals.

Examples

Demonstrates run logging, SQL metadata storage, object-storage artifacts, tracking-server proxying, and model-version retrieval.

What you can do after

Operate MLflow to record runs, retrieve metadata, manage artifacts, compare experiments, and govern model versions.

How it connects

Connects experiment tracking with metadata stores, artifact management, model registry, code reproducibility, model deployment, and MLMD.

liveBookmanning.com

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

Also explained in

• 

  liveBookmanning.com
  Machine Learning Platform EngineeringMLflow is an open-source platform for managing the machine learning lifecycle, including experiment tracking, model packaging, registry, and deployment.
• 

  liveBookmanning.com
  Generative AI in ActionMLflow is an open-source platform for managing machine-learning experimentation, reproducibility, model packaging, deployment, and lifecycle operations.

Compare how 3 books explain MLflow →

Related concepts

• experiment trackingused by · Experiment tracking is the systematic recording and organization of experiment configurations, metrics, artifacts, and results to support inspection, comparison, or reproducibility.
• Weights and Biasesalternative to · Weights & Biases is a machine-learning development platform for tracking experiments and managing runs, metrics, artifacts, models, and related workflows.
• model registryhas part · A model registry is a system for storing, organizing, versioning, and managing machine-learning models and their associated metadata throughout their lifecycle.
• platformkind of · A platform is a foundational computing system, environment, or online service that provides capabilities or infrastructure for building, running, delivering, or interacting with applications and services.
• MLflow clientused by · 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 Designing Deep Learning Systems: A guide for software engineers. The section links point to the parts of the book it draws on.