repurposing models

[anandbhattad]

Hi!

1. I am borrowing your models and repurposing them to another task. But I noticed that style tensors are initialized within the encoder block and are not being transferred to GPUs. Is it done on purpose?
   

   ALAE/net.py

   Line 309 in 42bcf2e

   styles = torch.zeros(x.shape[0], 1, self.latent_size)

2. This might be a naïve question. Can you please let me know what lod is/are? I come across them at multiple places in your code and couldn't understand what exactly they are.

Thanks for your time.

Best,
Anand