Batched probability distribution specification in Turing similar to Tensorflow Probability

In tensorflow probability there is this idea of batch_shape and event_shape while declaring a set of distribtuions. What is the equivalent of it in Turing.jl + Distributions.jl

For instance if I want to declare a batch of 3 multivariate normal distributions with event_shape 4, how will i do it in turing.jl + distributions.jl