orbit/template/dlt.py [660:692]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
            regressor_torch = torch.from_numpy(regressor_matrix).double()
            regression = torch.matmul(regressor_torch, regressor_beta)
            regression = regression.t()
        else:
            # regressor is always dependent with df. hence, no need to make full size
            regression = torch.zeros((num_sample, output_len), dtype=torch.double)

        ################################################################
        # Seasonality Component
        ################################################################

        # calculate seasonality component
        if self._seasonality > 1:
            if full_len <= seasonality_levels.shape[1]:
                seasonal_component = seasonality_levels[:, :full_len]
            else:
                seasonality_forecast_length = full_len - seasonality_levels.shape[1]
                seasonality_forecast_matrix = torch.zeros(
                    (num_sample, seasonality_forecast_length), dtype=torch.double
                )
                seasonal_component = torch.cat(
                    (seasonality_levels, seasonality_forecast_matrix), dim=1
                )
        else:
            seasonal_component = torch.zeros((num_sample, full_len), dtype=torch.double)

        ################################################################
        # Trend Component
        ################################################################

        # calculate level component.
        # However, if predicted end of period > training period, update with out-of-samples forecast
        if full_len <= trained_len:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



orbit/template/lgt.py [542:574]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
            regressor_torch = torch.from_numpy(regressor_matrix).double()
            regression = torch.matmul(regressor_torch, regressor_beta)
            regression = regression.t()
        else:
            # regressor is always dependent with df. hence, no need to make full size
            regression = torch.zeros((num_sample, output_len), dtype=torch.double)

        ################################################################
        # Seasonality Component
        ################################################################

        # calculate seasonality component
        if self._seasonality > 1:
            if full_len <= seasonality_levels.shape[1]:
                seasonal_component = seasonality_levels[:, :full_len]
            else:
                seasonality_forecast_length = full_len - seasonality_levels.shape[1]
                seasonality_forecast_matrix = torch.zeros(
                    (num_sample, seasonality_forecast_length), dtype=torch.double
                )
                seasonal_component = torch.cat(
                    (seasonality_levels, seasonality_forecast_matrix), dim=1
                )
        else:
            seasonal_component = torch.zeros((num_sample, full_len), dtype=torch.double)

        ################################################################
        # Trend Component
        ################################################################

        # calculate level component.
        # However, if predicted end of period > training period, update with out-of-samples forecast
        if full_len <= trained_len:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -