Computes the mode (value with the highest frequency) of a 1D numpy array. Parameters ---------- arr A numpy array representing th

Abstract property that returns model specific encoder settings that are used to initialize the encoders. Must return Tuple (input_chunk_lengt

Wrapper around model inference method

Wrapper around model inference method

Wrapper around model inference method

Return `True` if all the `forecasting_models` are probabilistic and fit the same distribution.

Returns the feature name mapping of the TFT model. Returns ------- Dict[str, str] The feature name mapping. For e

Helper function, used to make FFT model pickable.

Callback for when plotting.use_darts_style changes.

For TorchForecastingModels we use a strided inference dataset to avoid having to recreate trainers and datasets for each forecastable

Forecasts values for a certain number of time steps after the end of the series. TransferableFutureCovariatesLocalForecastingModel must implem

Convert boundaries to List

Has to be overwritten by the Likelihood objects supporting specifying a prior distribution on the outputs. If it returns None, no pri

Whether the model should be trained when calling a `fit*` method.

Gives the "per time step" metric and optional metric kwargs used to compute residuals / non-conformity scores.

Apply the scorer (sub) model scoring method on the series components

For each timestamp of the inputs: - the parameters of the considered distribution are fitted on the samples of the probabilistic time series

Compute score from confusion matrix components.

Check and set the likelihood. Only ClassProbability is supported for CatBoostClassifierModel.

Check and set the likelihood. Only ClassProbability is supported for LightGBMClassifierModel.

Returns the static covariates importance of the TFT model. The static covariate importances are calculated for the static inputs of the model

Specifies the supported SHAP methods.

Returns True if the model supports multi-output regression natively.

Checks if the forecasting model supports historical forecasts without retraining the model. By default, returns False. Needs

GlobalForecastingModel supports historical forecasts without retraining the model

Checks if the forecasting model supports a range index. Some models may not support this, if for instance they rely on underlying dates.

Prophet does not support integer range index.

Whether the model supports evaluation series passed to `fit()`.