Update VAAL to incorporate recent changes in framework
This MR updates VAAL in following ways
- Adds adapter to VAAL
- Adds python based config
- Create a yaml for VAAL
- Uses transforms that are similar to pretrain image classification
- Uses device along with cuda.
Currently the loss values for vae model is very high 1612180942224933122645950464.000. Not sure why this is happening on domain_net-painting.
Edited by Ameya Shringi