Let’s say, there are two plates and one is in another one. The first size is 10 and the second plate size is 5.
In this case, if my understanding is correct, should the batch_shape of the distribution be [10, 5]?
with pyro.plate(...,10, dim=-2)
some code
with pyro.plate(...,5, dim=-1)
z = pyro.sample(..., distribution, ...)