Element wise product with different dimension

I have a normal distribution with means and variances whose shape are [32, 1, 28, 28].
I want to get ten samples from the normal distribution as follows.
x = mean + variance * epsilon (epsilon is sampled from N(0,1) )

First, I call torch.normal(0, 1, size=[32, 1, 28, 28, 10]).
And then I want to product variance and epsilon, resulting in a [32, 1, 28, 28, 10] matrix.

How can I do it?