End-to-end time-series

An end-to-end tutorial to train a time-series model.

• Time-series forecasting with DLinear
  • Setup
  • Acknowledgements
• Distributed training of a DLinear time-series model
• Anatomy of a Ray Train job
  • The train configuration
  • Configuring persistent storage
  • Set up a training function
  • Set up the scaling config
  • Checkpointing configuration
  • Datasets
• DLinear model validation using offline batch inference
• Data ingest
• Online serving for DLinear model using Ray Serve
  • Set up the environment
  • Create a Ray Serve deployment
  • Load the model and start the service
  • Test the service
  • Production deployment considerations
    • Deploying to production