Python torch format to state_dict

How can i format my weights so I don't need to load state dict

m=torch.load(generator_path, map_location = device)
self.colorizer.generator.load_state_dict(m)
#self.colorizer.generator=m