Update README

Fixed bugs, updated ema_model to supress warnings, changed augmentation pipeline setup, unconditional generation support

scikit-image not skimage

skimage needed for augmentations

Add augmentations

Removed print debugs

New post hoc ema and augmentation training

Updated default config

update loss scaling config and use accelerator.save_state

removed clamp in model_a, fixed loss scaler issue, moved configs in configs folder

handle scaling hasattr

handle scaling

tweaks to config a

model config a

defaults update