Is it possible to observe a subset of variables created via shape=n?

Hello, I am trying to ‘observe’ two out of several bivariate Normal variables created together using the shape argument. The goal of conditioning these two variables is to break symmetry. Is there a good way to do this in PyMC?

with pm.Model() as model1:

    n_traits = 2        
    trait = pm.Normal('trait', mu=0, sigma=np.sqrt(1/np.sqrt(n_traits)), shape=(n_movies, n_traits))

    trait[101142] = [0,1]
    trait[4041] = [1,0]