orbit/template/dlt.py [560:610]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    def predict(
        self,
        posterior_estimates,
        df,
        training_meta,
        prediction_meta,
        include_error=False,
        **kwargs,
    ):
        """Vectorized version of prediction math"""
        ################################################################
        # Prediction Attributes
        ################################################################
        # n_forecast_steps = prediction_meta[PredictionMetaKeys.FUTURE_STEPS.value]
        start = prediction_meta[PredictionMetaKeys.START_INDEX.value]
        trained_len = training_meta[TrainingMetaKeys.NUM_OF_OBS.value]
        output_len = prediction_meta[PredictionMetaKeys.PREDICTION_DF_LEN.value]
        # this is only used in trend and seasonality generation
        full_len = prediction_meta[PredictionMetaKeys.END_INDEX.value]

        ################################################################
        # Model Attributes
        ################################################################
        model = deepcopy(posterior_estimates)
        for k, v in model.items():
            model[k] = torch.from_numpy(v)

        # We can pull any arbitrary value from teh dictionary because we hold the
        # safe assumption: the length of the first dimension is always the number of samples
        # thus can be safely used to determine `num_sample`. If predict_method is anything
        # other than full, the value here should be 1
        arbitrary_posterior_value = list(model.values())[0]
        num_sample = arbitrary_posterior_value.shape[0]

        # seasonality components
        seasonality_levels = model.get(
            SeasonalitySamplingParameters.SEASONALITY_LEVELS.value
        )
        seasonality_smoothing_factor = model.get(
            SeasonalitySamplingParameters.SEASONALITY_SMOOTHING_FACTOR.value
        )

        # trend components
        slope_smoothing_factor = model.get(
            BaseSamplingParameters.SLOPE_SMOOTHING_FACTOR.value
        )
        level_smoothing_factor = model.get(
            BaseSamplingParameters.LEVEL_SMOOTHING_FACTOR.value
        )
        local_trend_levels = model.get(BaseSamplingParameters.LOCAL_TREND_LEVELS.value)
        local_trend_slopes = model.get(BaseSamplingParameters.LOCAL_TREND_SLOPES.value)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



orbit/template/lgt.py [465:514]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    def predict(
        self,
        posterior_estimates,
        df,
        training_meta,
        prediction_meta,
        include_error=False,
        **kwargs,
    ):
        """Vectorized version of prediction math"""
        ################################################################
        # Prediction Attributes
        ################################################################
        # n_forecast_steps = prediction_meta[PredictionMetaKeys.FUTURE_STEPS.value]
        start = prediction_meta[PredictionMetaKeys.START_INDEX.value]
        trained_len = training_meta[TrainingMetaKeys.NUM_OF_OBS.value]
        output_len = prediction_meta[PredictionMetaKeys.PREDICTION_DF_LEN.value]
        full_len = prediction_meta[PredictionMetaKeys.END_INDEX.value]

        ################################################################
        # Model Attributes
        ################################################################
        model = deepcopy(posterior_estimates)
        for k, v in model.items():
            model[k] = torch.from_numpy(v)

        # We can pull any arbitrary value from teh dictionary because we hold the
        # safe assumption: the length of the first dimension is always the number of samples
        # thus can be safely used to determine `num_sample`. If predict_method is anything
        # other than full, the value here should be 1
        arbitrary_posterior_value = list(model.values())[0]
        num_sample = arbitrary_posterior_value.shape[0]

        # seasonality components
        seasonality_levels = model.get(
            SeasonalitySamplingParameters.SEASONALITY_LEVELS.value
        )
        seasonality_smoothing_factor = model.get(
            SeasonalitySamplingParameters.SEASONALITY_SMOOTHING_FACTOR.value
        )

        # trend components
        slope_smoothing_factor = model.get(
            BaseSamplingParameters.SLOPE_SMOOTHING_FACTOR.value
        )
        level_smoothing_factor = model.get(
            BaseSamplingParameters.LEVEL_SMOOTHING_FACTOR.value
        )
        local_trend_levels = model.get(BaseSamplingParameters.LOCAL_TREND_LEVELS.value)
        local_trend_slopes = model.get(BaseSamplingParameters.LOCAL_TREND_SLOPES.value)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -