From the documentation of pscore_match.pscore.PropensityScore() it is not clear that all categories have to binary in the form of zeros and ones (both the treatment and variables from the covariate pd.DataFrame). The documentation could be more explicit here.