This is a supportive material for the
- Blog Post Serve Many Forecasting Models with Databricks Model Serving at Once.
This code walks you through multiple approaches such as :
- How to train many independent forecasting models on top of Delta table using PandasUDF of Apache Spark.
- How to track each model using MLFlow and Delta tables.
- How to serialize all your models under a single MLFlow object.
- How to prepare your input data payload to be consumed by a single Endpoint.
- Data & AI summit talk Many Model Forecasting in Real-Time with Databricks Online Store
The dataset used for this code can be found here on the main Kaggle page.