is it possible to jointly train both global and local at same time ?

[wl082013]

I want to implement joint training for both local and global, but the train set (default) is with GlobalGenerator and no LocalEnhancer. Since the paper mentioned about joint training, so I wonder how to do it for my case. Could someone help explain how to jointly train both generators and how to fine-tune both without running different commands? thanks a lot !

[syfbme]

"python train.py --name test --label_nc 0 --netG local --niter_fix_global 1 --niter 1 --niter_decay 1" There are 2 stages including 2(niter+niter_decay ) epoch. After 1(niter_fix_global) epoch, both global and local will be trained together

[kex243]

if input is "-niter_fix_global 0 --netG local" only local generator will be used in traning? I have global G trained already and want to train local. Is there need to be trained for them together or I can train only local G if result from only global g is good already and req only minor improvments?

[daeunni]

@syfbme Hi, I wonder why i should set --niter 1 --niter_decay 1. do you have any idea?