Separate prior for each element of the batch.

[norpadon]

Why do you store separate prior mean and covariance for each sample within a batch?

glow-pytorch/glow/models.py

Line 199 in 487a6b1

nn.Parameter(torch.zeros([hparams.Train.batch_size // num_device,

It doesn't make sense to me. As I understand, you sample each element of the batch from its own distribution.