Components for individual features in LinearRegression and SparseLinearRegression

Tensortflow STS (Structural Time Series) allows to decompose the time series or forecast using sts.decompose_forecast_by_component method. For a model where one of the components is a LinearRegresion (or SparseLinearRegression) with multiple features, the decomposition returns only one series for this component - is there a way to break it down into effects of individual features?

Example:

def build_sts_model(observed_time_series, price_time_series):
    
    day_of_week_effect = sts.Seasonal(
        num_seasons=7, 
        observed_time_series=observed_time_series,
        name='day_of_week_effect')
    
    price_effect = sts.SparseLinearRegression(
                    design_matrix=tf.reshape(price_time_series, (-1, k)),
                    name='price_effect')
    
    model = sts.Sum([day_of_week_effect,
                     price_effect]
                    observed_time_series=observed_time_series)
    return model

...

#
# Decomposition:
#
forecast_component_dists = sts.decompose_forecast_by_component(
    demand_model,
    forecast_dist=demand_forecast_dist,
    parameter_samples=q_samples_demand_)

#
# returns one series fo all features combined,
# but how to get individual features?
#
forecast_component_dists['price_effect'].mean() 

Start with sts.decompose_forecast_by_component and the LinearRegression and SparseLinearRegression components used in the example. Determine whether decomposition can expose separate effects for each design-matrix feature rather than one combined series. Done means individual feature forecast series are available for the regression component.