I am trying to understand how to compute the posterior probability of a Gaussian process. I am using a Matern kernel

    ker = GPy.kern.Matern32(10, ARD=True)  

The mean is given by

K(xtest,xtrain)*K(xtrain,xtrain)^{-1}*ytrain

When trying to compute the second multiplication:

    chunk = np.linalg.solve(ker.K(xTrain,xTrain),yTrain)

I get a singularity error, and in effect the determinant of ker.K(xTrain,xTrain) is zero. What am I doing wrong here? Am I constructing the covariance matrices wrongly?