Bayesian calibration on GP model

Unfortunately, I haven’t been able to solve my issue yet. Has anyone already given pyMC variables to X (concatenated with known values) in the marginal_likelihood method in pyMC3 ?
Many thanks !