Faceswap Forum

Hi, is it possible to change the output size for one trainer in mid-train process or will it mess up the model? For example change dfaker output from 128 to 256? Can i expect better results in the endphase of training or is it not really helpful? Im aware of that a bigger output size needs more VRam and that you cannot change to a completly new trainer mid-train.

Thanks for your help and answers.

Tore wrote: ↑Sun Mar 28, 2021 12:05 pm

Hi, is it possible to change the output size for one trainer in mid-train process or will it mess up the model? For example change dfaker output from 128 to 256? Can i expect better results in the endphase of training or is it not really helpful? Im aware of that a bigger output size needs more VRam and that you cannot change to a completly new trainer mid-train.

Thanks for your help and answers.

Nope, it is not possible.
Once the setting you use to create a model, it gets locked to it & no
matter what you do, it will not affect that model.
Although, the next time you create a new model, these new settings
will be used for that new model.

The only things that you can UPDATE for an existing model are:-
1)Batch size
2)Learning Rate
3)Eye & Mouth multiplier
4)Warp, Flip & Color Augmentation

Thanks for your answer and the quick sumup.