Time Series Data Workload (Ray Train)

Learn how to build a time-series forecasting pipeline for NYC taxi demand using a sequence-to-sequence Transformer, predicting the next 24 hours from a week of history. You’ll also migrate the workflow from single-GPU PyTorch to distributed multi-node training on an Anyscale cluster with Ray Train V2.

• Time-Series workload pattern with Ray Train
• Imports
• PositionalEncoding and Transformer model
• Ray Train training loop (with teacher forcing)
• Resume training from checkpoint
• Inference helper — Ray Data batch predictor on GPU