MLflow: tracking, optional registration, consumption

MLflow: tracking, optional registration, consumption An architecture diagram generated by Archify. Tracking · Experiments, runs, traces · Covered by this documentation series (Parts 1-2) · Part 1 Tracking Experiments, runs, traces Part 1 Model Registry · Registered Models, aliases · Covered by this documentation series (Parts 1-2) · Part 2 Model Registry Registered Models, aliases Part 2 Serving · Out of scope · Architecture component Serving Out of scope optional registration consume URI / alias Covered by this documentation series (Parts 1-2) Legend Tracking Model Registry (core) Out of scope

Tracking (Part 1)

  • • Logs and queries experiment parameters, metrics, artifacts, models, and GenAI traces
  • • Covers experiments, runs, autologging, the MLflow 3 LoggedModel shift, and GenAI tracing

Model Registry (Part 2)

  • • Gives a model a stable, versioned identity independent of any one training run
  • • Covers Registered Models, Model Versions, aliases, and lineage
  • • Registration and alias updates do not imply automatic deployment