Use Input data

[beritHu]

I want to build a Gaussian process with 2-dim. Input data. the first row is the time and the second row is the patient ID. How can I predict the output of new time points using the already known data? For example, if I want to predict the output for additional time points for a patient where the information for the first n time points is already known?
If my model is called m1, I tried

gp_predict1 =gpflow.posteriors.GPRPosterior(m1.kernel, data_train, m1.likelihood.variance, m1.mean_function, precompute_cache=None)
gp_predict2.fused_predict_f(Xplot,full_cov=True)[0][:,1]

but even if some output of data_train should already be known, it isn't mapped on these data, eg. the known data is not taken into account.