Closed RewindL closed 1 year ago
line 419-422 is not used in my case, because my stylegan2 checkpoint (stylegan2-ffhq-config-f.pt) has no g_optim and d_optim.
You can just comment out line 419-422.
line 419-422 is not used in my case, because my stylegan2 checkpoint (stylegan2-ffhq-config-f.pt) has no g_optim and d_optim.
You can just comment out line 419-422.
OK, thanks for your answer.
In line 419-422 of pretrain_dualstylegan.py, two load_state_dict operations are executed to load dict of g_optim and d_optim from stylegan2 checkpoint. But I find a problem that g_optim in pretrain_dualstylegan.py contains different parameters from stylegan2 g_optim, and when I ran the code, I met
*** ValueError: loaded state dict contains a parameter group that doesn't match the size of optimizer's group
and I checked the lengths of parameters which are 74 and 110 respectively. So I have a question that is it reasonable to execute "g_optim.load_state_dict(ckpt["g_optim"])" ? Or there are something I missed or misunderstood ?