SundialOptions<T>

Options & Configuration · AiDotNet.Models.Options

Configuration options for Sundial (A Family of Highly Capable Time Series Foundation Models).

Inheritance: ModelOptions → RegressionOptions<T> → TimeSeriesRegressionOptions<T> → SundialOptions<T>

For Beginners

Sundial is a highly efficient forecasting model:

Key Innovation: Sundial outperforms Time-MoE (which has up to 2.4B params) with significantly fewer parameters, achieving a 4.71% average MSE reduction.

Architecture:

  • Decoder-only transformer (like GPT)
  • Patch-based input tokenization
  • Autoregressive generation for forecasting
  • Efficient scaling through architectural improvements

When to Use:

  • High-accuracy forecasting with moderate compute
  • When you need better accuracy than Time-MoE with fewer parameters
  • General-purpose time series forecasting across domains

How It Works

Sundial is a decoder-only time series foundation model that achieves state-of-the-art performance with fewer parameters than competing models like Time-MoE. It uses a GPT-style autoregressive architecture with patch-based tokenization.

Reference: “Sundial: A Family of Highly Capable Time Series Foundation Models”, 2025. https://arxiv.org/abs/2502.00816

Constructors

Constructor Summary
SundialOptions() Initializes a new instance with default values.
SundialOptions(SundialOptions<T> other) Initializes a new instance by copying from another instance.

Properties

Property Summary
FoundationModelSize ModelSize { get; set; } Gets or sets the model size variant.
bool UseFlashAttention { get; set; } Gets or sets whether to use flash attention for efficient computation.
double DropoutRate { get; set; } Gets or sets the dropout rate for regularization.
int ContextLength { get; set; } Gets or sets the context length (input sequence length).
int ForecastHorizon { get; set; } Gets or sets the forecast horizon (prediction length).
int HiddenDimension { get; set; } Gets or sets the hidden dimension of the transformer.
int IntermediateSize { get; set; } Gets or sets the intermediate size for the feed-forward network.
int NumHeads { get; set; } Gets or sets the number of attention heads.
int NumLayers { get; set; } Gets or sets the number of transformer layers.
int NumQuantiles { get; set; } Gets or sets the number of quantiles for probabilistic forecasting.
int PatchLength { get; set; } Gets or sets the patch length for input tokenization.