How to train the StyleGAN2Generator used as prior knowledge if i want to make some modifications

[NJUPT-IPR-ZhuWeihong]

Ur work is amazing! I want to make some modifications in the StyleGAN2Generator network.Due to modifications above,the pretrained models of StyleGAN2Generator cannot be used any more.What's more,as ur paper mentioned, generative facial prior gives rise to performance improvement.
Therefore, I have to train StyleGAN2Generator solely to achieve better performance.I am wondering if i could just modify the official Pytorch StyleGAN2 code(the train file, mainly the network architecture) and the model file will adapt to GFPGAN. Of course i modify GFPGAN(the train file and the inference file) at the same time.
Is there any tips or warnings?
It will be appreciated a lot if i could get ur guidance.
Whatever, thanks for ur amazing work in FSR again!

[NJUPT-IPR-ZhuWeihong]

I found a yml file in basicsr\options\train\StyleGAN, named train_StyleGAN2_256_Cmul2_FFHQ.yml.Maybe it will work?
I don't know.Trying...

[zzuyb]

I found a yml file in basicsr\options\train\StyleGAN, named train_StyleGAN2_256_Cmul2_FFHQ.yml.Maybe it will work? I don't know.Trying...

Can it be used normally？

[NJUPT-IPR-ZhuWeihong]

I found a yml file in basicsr\options\train\StyleGAN, named train_StyleGAN2_256_Cmul2_FFHQ.yml.Maybe it will work? I don't know.Trying...

Can it be used normally？

yes,just modify some parameters according to ur requirements