Some hierarchical modeling techniques require evaluating different statistics on the same sample of the posterior. Currently, there is no seed to set for the sample function.
I confirmed the code is using numpy for sampling. As a temporary fix, you can set np.random.seed(n) before running the sample function each time if you need the same samples.
Some hierarchical modeling techniques require evaluating different statistics on the same sample of the posterior. Currently, there is no seed to set for the sample function.