Can I get channel contribution inference from out of sample prediction

[kkgg]

There are only y values in the result. I need something similar to mmm.compute_mean_contributions_over_time() - here I can obtain contribution of each channel, yearly seasonality and intercept, but it works for the train data.