def _load_optimizer_state()

in improved_diffusion/train_util.py [0:0]

• 11 lines of code
• 3 McCabe index (conditional complexity)

    def _load_optimizer_state(self):
        main_checkpoint = find_resume_checkpoint() or self.resume_checkpoint
        opt_checkpoint = bf.join(
            bf.dirname(main_checkpoint), f"opt{self.resume_step:06}.pt"
        )
        if bf.exists(opt_checkpoint):
            logger.log(f"loading optimizer state from checkpoint: {opt_checkpoint}")
            state_dict = dist_util.load_state_dict(
                opt_checkpoint, map_location=dist_util.dev()
            )
            self.opt.load_state_dict(state_dict)