version 1.0
train script: train_script.py sample script: sample_script.py
still remains trouble
precision default to fp16, SGD optimization, xformers, image_size & channel dim set to 256.(OOM if set to 512...) Why does GPU memory still exceed 20G?