NV-Tesseract-Forecasting Overview

Description:

NVIDIA NV-Tesseract-Forecasting provides forecasting functionality via NVIDIA's novel DARR (Domain-Aware Representation and Retrieval) which is a context-enhanced forecasting mode that combines direct model predictions with pattern retrieval from historical data. The model leverages MOMENT encoders with a trainable forecasting head for accurate long-horizon time series predictions.

This model is for research and development only.

License/Terms of Use:

GOVERNING TERMS: Use of this model is governed by the Apache-2.0 License

Deployment Geography:

Global

Use Case:

Companies, organizations, research hubs looking to do long/short-horizon forecasting on temporal data.

Release Date:

HuggingFace 04/07/2026 via https://huggingface.co/nvidia/nv-tesseract-forecasting

Reference(s):

Memory-Augmented Forecasting: Scalability and Generalization Across Temporal Domains
Beyond MAE: Measuring Forecast Reliability with Temporal Dependence-Aware Error (TDE)
MOMENT: A Family of Open Time-series Foundation Models
TimeRAF: Retrieval-Augmented Foundation model for Zero-shot Time Series Forecasting

Model Architecture:

Architecture Type: MOMENT-based Transformer with trainable forecasting head

Network Architecture: Transformer Encoder (MOMENT-1-large backbone)

Number of model parameters: ~340 million (base encoder) + trainable forecasting head

The NV-Tesseract-Forecasting is a Transformer-based model for time series forecasting. It uses MOMENT encoders with frozen encoder and embedder weights, while training only the forecasting head. The model supports DARR (Domain-Aware Representation and Retrieval) mode, which enhances predictions by combining direct forecasts with kNN-based pattern retrieval from historical context data.

Input:

Input Type(s): Tabular numeric
Input Format(s): Tabular Pandas DataFrame or CSV/JSON
Input Parameters: 2D
Other Properties Related to Input: Pre-Processing Needed
Forecasting: Contains timestamp column and one or more numeric value columns. Supported sequence lengths: 256, 512, 1024, 2048.

Output:

Output Type(s): Tabular numeric
Output Format: Tabular Pandas DataFrame
Output Parameters: 2D
Other Properties Related to Output: Post-Processing Needed
Forecasting: Contains timestamp and predicted value columns for the specified forecast horizon. Supports autoregressive extension for horizons beyond the model's native capability.

Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA's hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions.

Software Integration:

Runtime Engine(s):
PyTorch

Supported Hardware Microarchitecture Compatibility:
NVIDIA Ampere
NVIDIA Hopper

[Preferred/Supported] Operating System(s):
Linux

The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.

Model Version(s):

NV-Tesseract-Forecasting

Training & Testing Datasets:

80/20 split per dataset.

Data Modality: Other: Numeric time series

Training Data Size: 3 million data points

Monash Time Series Forecasting Archive

Data Collection Method by dataset [Hybrid: Synthetic, Sensors, Human]
Labeling Method by dataset [Hybrid: Synthetic, Sensors, Human]
Properties: A diverse collection of 30 primary time series datasets used for global forecasting model evaluation. Includes missing value variations.

Evaluation Datasets:

ECL (Electricity Load Diagrams)

Data Collection Method by dataset [Human,Sensors]
Labeling Method by dataset [Human,Automated]
Properties: Electricity consumption data from 370 clients recorded every 15 minutes from 2011 to 2014, widely used for long-term forecasting benchmarks.

Traffic

Data Collection Method by dataset [Humans,Sensors]
Labeling Method by dataset [Humans,Automated]
Properties: Road occupancy rates measured by sensors on San Francisco Bay area freeways, commonly used for traffic flow forecasting evaluation.

ETTh (Electricity Transformer Temperature)

Data Collection Method by dataset [Humans,Sensors]
Labeling Method by dataset [Humans,Automated]
Properties: Electricity transformer temperature data collected over two years at hourly intervals, including oil temperature and six power load features (ETTh1, ETTh2).

ILI (Influenza-Like Illness)

Data Collection Method by dataset [Healthcare Surveillance]
Labeling Method by dataset [Human]
Properties: Weekly data from U.S. Centers for Disease Control describing the ratio of patients with influenza-like illness to total patients.

Inference:

Engine: PyTorch, Transformer Engine
Test Hardware:

• A100 (8 GPUs; each is 74 GB)
  
• H100 (8 GPUs; each is 80 GB)

Ethical Considerations:

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

Please report security vulnerabilities or NVIDIA AI Concerns here.