NVIDIA releases Kumo Forecast, a time-series model with retrieval and interpretability
AINVIDIA's Kumo Forecast, formerly NV-Tesseract Forecasting, is a multivariate time-series model built on the MOMENT-1-large backbone with an approximately 340-million-parameter base encoder and a trainable forecasting head. It adds Domain-Aware Representation and Retrieval (DARR), which blends direct predictions with patterns retrieved from historical context data, and an interpretability framework that produces per-horizon explanations.
Source details
Hugging Face Blog · huggingface.coPublished · added here
Original title: NVIDIA Kumo Forecast Combines Pretrained Time-Series Representations with Historical Pattern Retrieval and Interpretability