A batch of samples from a MultivariateNormal?

My question is two fold:

  1. Is that possible to produce multiple samples from a multivariateNormal?
    For instance for a latent space of size 10 and num_samples, I do something like this:
sampler = torch.distributions.multivariate_normal.MultivariateNormal(mean.repeat(num_samples,1,1),torch.eye(10))
samples = sampler.sample()

Is this correct?

  1. What if each sample has a different mean and var?