Hi there, I found a solution in my case to this problem. In my case, I realized that the following line is causing the issue:
betas = at.reshape(at.tile(alphas.T,num_respondents), (num_respondents, num_items)) + at.dot(betas_init,L_sigma)
I had to do the above calculation because I was taking samples from a standard multivariate normal and then needed to do the above transformation. To avoid that, I took samples from a multivariate normal with mean of “alphas” and covariance of “L_sigma.dot(L_sigma.T)” and then the above line was not necessary anymore.