Method checkpoint

scripts/diffusers_textual_inversion_2.py:488–512

(self, step, text_encoder, save_samples=True, path=None)

    @torch.no_grad()
    def checkpoint(self, step, text_encoder, save_samples=True, path=None):
        print("Saving checkpoint for step %d..." % step)
        with torch.autocast("cuda"):
            if path is None:
                checkpoints_path = f"{self.output_dir}/checkpoints"
                os.makedirs(checkpoints_path, exist_ok=True)

            unwrapped = self.accelerator.unwrap_model(text_encoder)

            # Save a checkpoint
            learned_embeds = unwrapped.get_input_embeddings().weight[
                self.placeholder_token_id
            ]
            learned_embeds_dict = {
                self.placeholder_token: learned_embeds.detach().cpu()
            }

            filename = "%s_%d.bin" % (slugify(self.placeholder_token), step)
            if path is not None:
                torch.save(learned_embeds_dict, path)
            else:
                torch.save(learned_embeds_dict, f"{checkpoints_path}/{filename}")
                torch.save(learned_embeds_dict, f"{checkpoints_path}/last.bin")
            del unwrapped
            del learned_embeds

    @torch.no_grad()
    def save_samples(