Models · Updated 21 Oct, 06:21 pm IST
NVIDIA releases Kumo Forecast time‑series model with retrieval and interpretability
Why it matters for readers: It demonstrates a method that mixes learned patterns and historical lookups to make forecasts more accurate and explainable.
- Kumo Forecast uses a MOMENT‑1‑large backbone (about 340 million parameter encoder) plus a trainable forecasting head.1
- The model supports an optional Domain‑Aware Representation and Retrieval (DARR) mode that retrieves patterns from historical context and blends them with direct model predictions.1
- The library provides a DataFrame‑first API, configurable forecast horizons, optional historical context for retrieval, and horizon‑specific interpretability without modifying the forecaster.1
- Model code and weights are published on Hugging Face under Apache 2.0 as NVIDIA/Kumo‑TS and nvidia/Kumo‑Forecast respectively.1
Get a brief like this every morning
Uzha reads hundreds of sources and gives you the stories that matter for your work, with every source linked. Free.
Get started