End-to-end distributed XGBoost

An end-to-end tutorial to train a distributed XGBoost model and perform offline batch inference and online serving at scale.

• Overview
• Distributed training of an XGBoost model
  • Dataset preparation
  • Data preprocessing
  • Model training with XGBoost
  • Model registry
  • Scaling strategies
• Model validation using offline batch inference
  • Validating the model using Ray Data
  • Observability
  • Summary
• Scalable online XGBoost inference with Ray Serve
  • Loading the model
  • Creating a Ray Serve deployment
  • Preparing test data
  • Running the service
  • Querying the service