Method save_checkpoint

funasr/train_utils/trainer_ds.py:171–412

Saves a checkpoint containing the model's state, the optimizer's state, and the scheduler's state at the end of the given epoch. This method is intended to be called at the end of each epoch to save the training progress. Args: epoch (int): The epoch num

(
        self,
        epoch,
        step=None,
        model=None,
        optim=None,
        scheduler=None,
        scaler=None,
        step_in_epoch=None,
        **kwargs,
    )

        self.effective_save_name_excludes = effective_save_name_excludes

    def save_checkpoint(
        self,
        epoch,
        step=None,
        model=None,
        optim=None,
        scheduler=None,
        scaler=None,
        step_in_epoch=None,
        **kwargs,
    ):
        """
        Saves a checkpoint containing the model's state, the optimizer's state,
        and the scheduler's state at the end of the given epoch. This method is
        intended to be called at the end of each epoch to save the training progress.

        Args:
            epoch (int): The epoch number at which the checkpoint is being saved.
        """
        if self.use_ddp or self.use_fsdp:
            dist.barrier()
        step_in_epoch = None if step is None else step_in_epoch
        if self.use_deepspeed:

            logging.info(f"Save checkpoint: {epoch}, rank: {self.local_rank}\n")
            # self.step_or_epoch += 1
            state = {
                "epoch": epoch,
                # "state_dict": model.state_dict(),
                # "optimizer": optim.state_dict(),
                # "scheduler": scheduler.state_dict(),
                "saved_ckpts": self.saved_ckpts,
                "val_acc_step_or_epoch": self.val_acc_step_or_epoch,
                "val_loss_step_or_epoch": self.val_loss_step_or_epoch,
                "best_step_or_epoch": self.best_step_or_epoch,
                "avg_keep_nbest_models_type": self.avg_keep_nbest_models_type,
                "step": step,
                "step_in_epoch": step_in_epoch,
                "data_split_i": kwargs.get("data_split_i", 0),
                "data_split_num": kwargs.get("data_split_num", 1),
                "batch_total": self.batch_total,
                "train_loss_avg": kwargs.get("train_loss_avg", 0),
                "train_acc_avg": kwargs.get("train_acc_avg", 0),
            }
            step = step_in_epoch
            if hasattr(model, "module"):
                state["state_dict"] = model.module.state_dict()

            if scaler:
                state["scaler_state"] = scaler.state_dict()
            # Create output directory if it does not exist
            os.makedirs(self.output_dir, exist_ok=True)
            if step is None:
                ckpt_name = f"model.pt.ep{epoch}"
            else:
                ckpt_name = f"model.pt.ep{epoch}.{step}"
            filename = os.path.join(self.output_dir, ckpt_name)