Raises `ValueError` if `self._parallel_params` specifies a `key` in `self._fixed_params` that should be distributed, but `len

Historical forecast can be optimized only if `retrain=False`.

Checks if the conditions for custom conditional seasonalities are met. Each custom seasonality that has a `condition_name` other than

Raises an error if the start index (position) is not within the series.

Checks if `anomalies` contains only univariate series when the scorer has the parameter 'is_univariate' set to True. 'is_univariate'

Compute the central TimeSeries for a component.

Compute the low and high quantile TimeSeries for confidence intervals. Returns ------- tuple[TimeSeries, TimeSeries] A tuple of (

Computes a confusion matrix using numpy for two np.arrays `y_true` and `y_pred`. Parameters ---------- y_true The true labels.

Creates the bounds for the inference dataset based on the input series and whether it is for training or not.

Function creating the lower frequency dataframe out of a higher frequency dataframe.

Create work dir and model dir

Differentiate the series self.d times

a/b where the resulted NaN or Inf are replaced by 0.

Sequentially encodes the index of all input target/covariates TimeSeries with the corresponding `encoder_method`.

Computes a score/metric between anomaly scores or binary predicted anomalies against true anomalies. Parameters ---------- anomalies

Extends a `time_index` of frequency `freq` such that it now ends at time `new_end`; the fastest way to do this is actually to create a new ti

Helper function called by `_create_lagged_data_by_moving_window` that reshapes the `windows` formed by `strided_moving_window` from the s

Extracts sample weights values from the time intersection with the target labels.

Return the indices of the first rows containing finite values starting from the start and the end of the first dimension of the ndarray. Para

Function that fit the model. Deriving classes can override this method for adding additional parameters (e.g., adding validation data

Train the fittable scorers using model forecasts

Subclasses can override this method to format the samples and labels before fit and predict.

Generates `n` new dates after the end of the specified series

Generates `n` new dates after the end of the specified series

Returns a pandas DataFrame in the format required for Prophet.predict() with `n` dates after the end of the fitted TimeSeries

The returned index depends on the following cases: case 1 (steps_ahead_start >= 0 and steps_ahead_end is None or <= 1) the comple

Returns the indices and column names of the categorical features in the regression model. Steps: 1. Get the list of features

Get the path to a file either from a local directory or by downloading it from HuggingFace. Parameters ---------- filename

Returns list with the frequency of all specified (i.e. non-`None`) `series`.

Based on the boundaries of the forecastable time index, generates the boundaries of each covariates using the lags. For TimeSeries with a Ra

Get model description string of structure `model_name`(`model_param_key_value_pairs`). Parameters ---------- include

Creates generator of dictionaries containing fixed parameter values (i.e. attributes defined in the child-most class). Those fixed pa

Finds a valid historical forecast start point within either `series` or `historical_forecasts_time_index` (depending on whether `historical_foreca

Computes the OLS residuals for predicting the target series from `future_covariates`.

Returns a deterministic or probabilistic numpy array from the values of a time series of shape (times, components, samples / quantiles).

Return the kwargs dict without the arguments unsupported by the model method. Raise a warning if some argument are not supported and an exce

Initializes model and trainer based on examples of input/output tensors (to get the sizes right):

Check if the specified function is a method. Parameters ---------- func the function to inspect. Returns ------- boo

Returns dict with feature names and (start, end) index ranges. The features are (past target, future target, past cov, future past cov, histo

Return all observed/inferred frequencies of a ``pandas.Index`` (an integer-valued index). The inferred frequencies are given by all unique di

Packs each provided input into a list (or str in case of sample weight).

Average across the components after grouping by likelihood parameter, rename components

Returns the distribution parameters, obtained from the raw model outputs (e.g. applies softplus or sigmoids to get parameters in the

Reduce the sample dimension of the forecasting models predictions

Process the boundaries argument and perform some sanity checks Parameters ---------- lower_bound_name Na

Shared input validation and parameter preparation for plot() and plotly().

Checks that the `historical_forecasts` have the correct format based on the input `series` and `last_points_only`. If all checks have passed, it c

Processes module input batch. Converts output of a dataset into a tuple of tensors (past target + past cov + historic future cov (con

Checks input series and coverts series and covariates in `kwargs` to sequences.

Processes the historical forecastable time index bounds (earliest, and latest possible prediction start points). Parameters ----------

Extracts the time index, and optionally adds some time steps after the end of the index, and/or converts the index to another time zone.

Validates the validation set and generates/adds the required encodings.

overwrite parent classes `_produce_predict_output` method

Reversible instance normalization.

Verify that the assumptions for likelihood parameters prediction are verified: - Probabilistic models fitted with a likelihood - `num_

Performs sanity check on the two given inputs Checks if the two inputs: - type is Darts Timeseries - have the same number of comp

This method acts on `TimeSeries` inputs. It performs sanity checks, and sets up / returns the datasets and additional inputs required for trai

This method acts on `TorchTrainingDataset` inputs. It performs sanity checks, and sets up / returns the trainer, model, and dataset loaders re

Sets up a PyTorch-Lightning trainer (if not already provided) for training or prediction.

Computes the slice intersection of all series sequences. Raises an error if the intersection is empty for any of the sequences.

Checks the validity of an add_seasonality() call and stores valid calls. As the actual model is only created at fitting time, and seasonalitie

Average across the components, keep n_samples, rename components

performs a training or validation step

Transforms the static covariates of a `series` if `method = 'transform'`, and inverse transforms the static covariates of a `series`

Based on the Forecasting class and the training_sample attribute, update the uses_[past/future/static]_covariates attributes.

Dataset output dtype checks. Checks that all dataset output arrays have the same dtype, and whether the dtype matches the one of the

Check the base requirements for `min_covariates_lag` and `max_covariates_lag`: - both must either be None or an integer - min_covariat

Simple check if user supplied/did not supply covariates as done at fitting time.

Verify that any non-None covariates comply with the model type.

Verify that any non-None covariates comply with the model type.

Some sanity checks on the input, the high_freq and low_freq arguments are mutually exclusive

Extends the active cells in the column by the range (start,end). Ranges smaller than the current one are ignored. Note (1, m+1), not (

Creates a univariate, autoregressive TimeSeries whose values are calculated using specified coefficients `coef` and starting values `start_va

r"""Compute error values that the model produced for historical forecasts on (potentially multiple) `series`. If `historical_forecasts` are p