Challenge 2: How quickly can you start training?
< Previous Challenge - Home - Next Challenge >
Introduction
The objective of this challenge is to create an AutoML forecasting model for the next 30 days using the dataset from the first challenge. Instead of a point forecast we want you to output quantile forecasts so you can fine tune your results based on your business use case. We already provided you the skeleton of the function but you need to fill it in.
Success Criteria
- There’s a new model trained successfully on VertexAI
- Code changes are needed for this challenge
Tips
- Be careful on which metric you choose as an optimization objective. Quantile predictions require a specific parameter.
- Training will take ~2.5 hours to complete.
Learning Resources
Training documentation is here & tabular training kick-off documentation is here. Read optimization objectives for forecast models here.