← Back
Hugging Face Blog

Ensemble forecasting with IBM Granite time-series models

•5 min read•
#inference
Level:Intermediate
For:AI Engineers
✦TL;DR

The Granite Timeseries Ensemble bundles four pre‑trained Granite forecasting models into a single recipe that delivers median and quantile predictions without any retraining or fitted ensemble weights. The default Probabilistic Ensemble uses uniform probability pooling to merge each model’s quantile forecasts (0.1‑0.9) into a combined predictive distribution, while an alternative IQR Ensemble offers adaptive weighting. Engineers can pull the recipe directly from Hugging Face via the Granite TSFM library, enabling rapid deployment of a diversified, out‑of‑the‑box time‑series solution.

⚡ Key Takeaways

  • Four pre‑trained Granite models are combined in the default ensemble recipe.
  • The ensemble uses a Probabilistic Ensemble (uniform probability pooling) to merge quantile forecasts.
  • Quantile outputs range from 0.1 to 0.9, providing an 80 % nominal coverage interval.
  • Integration is done through the Granite TSFM library’s API, accessing `result.prediction_outputs` for point forecasts and `result.quantile_outputs` for distributional forecasts.
  • No model retraining or ensemble weight fitting is required, but the four member checkpoints are downloaded at runtime, so the first run incurs extra latency.
💡 Why It Matters

This approach lets production engineers obtain robust, diversified forecasts instantly, reducing model selection risk and eliminating the overhead of training or tuning ensemble weights.

✅ Practical Steps

  1. Run `pip install granite-tsfm` and load the recommended four‑model recipe from Hugging Face as shown in the example.
  2. Access `result.prediction_outputs` for the median forecast and `result.quantile_outputs` (0.1‑0.9) for uncertainty intervals in your application.

Want the full story? Read the original article.

Read on Hugging Face Blog ↗

More like this

Helping AI text detectors keep up with changing conditions

MIT News AI•#llm

Hugging Face and OpenRouter join forces to make open-weight models easy to run

Hugging Face Blog•#inference

Activation and Floating-Point Quantization

Machine Learning Mastery•#inference

New research could lead to faster, trustworthy AI models for climate prediction

MIT News AI•#inference

EXPLORE AI NEWS

Daily hand-picked stories on LLMs, RAG, agents and production AI — curated for engineers who ship.

BROWSE NEWS

GET THE WEEKLY DIGEST

Join engineers getting the Monday signal-over-noise AI breakdown. No spam, unsubscribe anytime.

LEARN AI ENGINEERING

Curated courses, research papers, repos and tutorials built for engineers leveling up in AI.

START LEARNING