@martinmodrak: Thank you so much for your help. I figured out the issue by putting the entire problem and dataset onto Amazon’s AWS server and run on there. And yes, yT is necessarily >0 in my problem.
By the way, could you please help take a look at this model in the other post I just posted (Sampling error: Unrecoverable error evaluating the log probability at the initial value)?? I wonder if you can kindly let me know if my marginalization process was correct, in the sense that from the log_lambda I obtained, I somehow could get the sampled values for the posterior distribution given by (Poisson x Normal) = Prior x Likelihood (model I used is Hierarchical Bayesian).