loglkhs = multi.log_prob(exp_values * shots) - multi.log_prob(sim_values * shots)