ChronosBoltOptions<T>

Options & Configuration · AiDotNet.Models.Options

Configuration options for Chronos-Bolt (Fast Non-Autoregressive Time Series Forecasting).

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

For Beginners

Chronos-Bolt trades autoregressive generation for speed:

Key Difference from Chronos v1/v2:

  • Chronos v1/v2: Generates one token at a time (autoregressive, slow)
  • Chronos-Bolt: Generates all forecast steps at once (non-autoregressive, fast)

Architecture:

  • Encoder: Processes the input context
  • Decoder: Directly outputs all forecast quantiles in one pass
  • No autoregressive loop = much faster inference

When to Use:

  • When you need fast inference (production/real-time)
  • When Chronos v1/v2 is too slow for your use case
  • When you need quantile forecasts (uncertainty estimates)

How It Works

Chronos-Bolt is part of the Amazon Chronos family but uses an encoder-decoder architecture with direct quantile forecasting (non-autoregressive), making it significantly faster than the autoregressive Chronos v1/v2 models while maintaining competitive accuracy.

Reference: Part of Chronos family, Amazon, Nov 2024.

Constructors

Constructor Summary
ChronosBoltOptions() Initializes a new instance with default values.
ChronosBoltOptions(ChronosBoltOptions<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.
double DropoutRate { get; set; } Gets or sets the dropout rate.
double EndLearningRate { get; set; } Gets or sets the final learning rate the cosine decay winds down to at TotalSteps.
double LearningRate { get; set; } Gets or sets the base (peak) learning rate for the built-in Adam optimizer with linear warmup.
int ContextLength { get; set; } Gets or sets the context length.
int DecoderHiddenDim { get; set; } Gets or sets the decoder hidden dimension.
int EncoderHiddenDim { get; set; } Gets or sets the encoder hidden dimension.
int ForecastHorizon { get; set; } Gets or sets the forecast horizon.
int NumDecoderLayers { get; set; } Gets or sets the number of decoder layers.
int NumEncoderLayers { get; set; } Gets or sets the number of encoder layers.
int NumHeads { get; set; } Gets or sets the number of attention heads.
int NumQuantiles { get; set; } Gets or sets the number of quantiles for direct quantile forecasting.
int PatchLength { get; set; } Gets or sets the patch length for input tokenization.
int TotalSteps { get; set; } Gets or sets the total number of scheduler steps (batches) over which the learning rate decays from LearningRate down to EndLearningRate.
int WarmupSteps { get; set; } Gets or sets the number of warmup steps (batches) over which the learning rate ramps from a tiny initial value up to LearningRate.