Foundation Models API

ChronosForecaster

ChronosForecaster(modelId="amazon/chronos-bolt-small", device="cpu")

Amazon Chronos-2 wrapper.

Methods

• fit(y) → self (stores context, no training)
• predict(steps) → (predictions, lower, upper)
• predictQuantiles(steps, quantileLevels) → ndarray
• predictBatch(series_list, steps) → list of tuples

TimesFMForecaster

TimesFMForecaster(modelId="google/timesfm-2.5-200m-pytorch")

Google TimesFM 2.5 wrapper.

Methods

• fit(y) → self
• predict(steps) → (predictions, lower, upper)
• predictWithCovariates(steps, dynamicNumerical, dynamicCategorical) → tuple

NeuralForecaster

NeuralForecaster(architecture="nbeats", **kwargs)

NeuralForecast wrapper.

Convenience Classes

• NBEATSForecaster(**kwargs)
• NHITSForecaster(**kwargs)
• TFTForecaster(**kwargs)

Availability Flags

from vectrix import CHRONOS_AVAILABLE, TIMESFM_AVAILABLE, NEURALFORECAST_AVAILABLE

Boolean flags indicating whether optional dependencies are installed.