Can this code realize training a single model and predicting multiple X.X SR images?

[jeanfang]

Hi @smallsunsun1 , thanks for sharing your code! Nice work!
I have some questions: "D.scale" sets the size of scale factor, while training meta-SR, you said " training_process split into several period, scale ranges in 2,3,4" , how to realize this step exactly? Will this code train a single model to test arbitrary upscale SR, and if it's not a single model, how to do the X.X upscale tests ? Hoping for you answer and thanks for your time!

[smallsunsun1]

code in util.py deal with it. if you select RDN model, D.scale will use. if you select Meta-SR model, D.scale if meaningless and will use a random number for scale rather than D.scale.

[suke27]

how can i training multi scales during traing, e.g. training 1.1 to 4.0 by step 0.1