Open ggamaz opened 1 week ago
You can load pre-trained models in the following ways:
try:
model.load_state_dict(torch.load(opt.model_path, map_location='cpu'), strict=True)
except:
from collections import OrderedDict
from copy import deepcopy
state_dict = torch.load(opt.model_path, map_location='cpu')['model']
pretrained_dict = OrderedDict()
for ki in state_dict.keys():
pretrained_dict[ki[7:]] = deepcopy(state_dict[ki])
model.load_state_dict(pretrained_dict, strict=True)
Thanks for your great work sincerely! I wonder what's the difference between the NPR.pth and model_epoch_last_3090.pth,I want to load the NPR.pth, but there is an error, which looks like about model structure, i donot know how to do