IBM releases Granite Timeseries Ensemble for time-series forecasting

Summary

IBM Research released Granite Timeseries Ensemble, a tool combining four open pre-trained time-series AI models into a single interface. The release provides two combination recipes, a Probabilistic Ensemble and an IQR Ensemble, which return median and quantile forecasts without requiring model retraining or fitted ensemble weights.

What changed

• Combines four open Granite time-series AI models into an extensible forecasting interface
• Provides a Probabilistic Ensemble recipe using uniform probability pooling with fixed weights
• Provides an adaptive IQR Ensemble recipe that weights members based on forecast spread
• Allows loading the default recipe directly from Hugging Face using Python

Availability

“Install Granite TSFM and load the recommended four-model recipe directly from Hugging Face:”

Source