Feature Flag (in ML)
A mechanism to enable or disable ML model features or entire models in production without redeploying code.
Overview
Feature flags in ML systems allow teams to toggle model versions, enable experimental features for subsets of users, and perform gradual rollouts. They decouple model deployment from model release, reducing risk.
Use Cases
Common uses include A/B testing model variants, gradual traffic shifting between model versions, emergency model rollback, and enabling different models for different user segments or geographies.