Can a GP prior be reshaped for function input?


I have been trying to use a GP as an input for a function which is used for the likelihood. Some pseudo code would be:

likelihood = pm.Normal('obs', mu=func, sigma=sd, observed=obs)

But the function only take in a 4-D array so i have reshape f since gp.prior only takes in arrays of the form (length(array),1). Luckily that is possible to do inside of the model

f = f.reshape((1,1,len(X),1))

and the sampler does seem to accept this. My question is does pymc “know” what is going on with the gp prior after i reshape? Will it still sample f (f_rotated) after i have reshaped it?