windelvin
Results
1
comments of
windelvin
> Another quick question, Have you tried using 1024*1024 resolution to train the model, Does lr=1e-4 still work? Or the dataset needs to be expanded?