在 lightning.pytorch Trainer 中使用回调时出现“父错误”

Question

我正在尝试在 colab 笔记本中复制 here 中显示的多模式转换器教程。然而，这是一个相对较旧的脚本，lightning.pytorch 发生了显着变化。我已将它适应新的闪电，当我在 Trainer 中删除回调参数时它会运行，但是当我添加回调时它会抛出以下错误：

/usr/local/lib/python3.9/dist-packages/lightning/pytorch/utilities/model_helpers.py in is_overridden(method_name, instance, parent)
     32             parent = pl.Callback
     33         if parent is None:
---> 34             raise ValueError("Expected a parent")
     35 
     36     from lightning_utilities.core.overrides import is_overridden as _is_overridden

ValueError: Expected a parent

这里是调用trainer的fit方法

def fit(self):
    # print(self)
    self._set_seed(self.hparams.get("random_state", 42))
    # self.trainer = pl.Trainer()
    self.trainer = pl.Trainer(callbacks=self.trainer_params)

和构建回调列表的 _get_trainer_params 方法

def _get_trainer_params(self):
        checkpoint_callback = pl.callbacks.ModelCheckpoint(
            dirpath=self.output_path,
            # filepath=self.output_path,
            monitor=self.hparams.get(
                "checkpoint_monitor", "avg_val_loss"
            ),
            mode=self.hparams.get(
                "checkpoint_monitor_mode", "min"
            ),
            verbose=self.hparams.get("verbose", True)
    )

    early_stop_callback = pl.callbacks.EarlyStopping(
        monitor=self.hparams.get(
            "early_stop_monitor", "avg_val_loss"
        ),
        min_delta=self.hparams.get(
            "early_stop_min_delta", 0.001
        ),
        patience=self.hparams.get(
            "early_stop_patience", 3
        ),
        verbose=self.hparams.get("verbose", True),
    )

    trainer_params = [
        checkpoint_callback,
        early_stop_callback,
        self.output_path,
        self.hparams.get(
            "accumulate_grad_batches", 1
        ),
        self.hparams.get("n_gpu", 1),
        self.hparams.get("max_epochs", 100),
        self.hparams.get(
            "gradient_clip_value", 1
        )
    ]

再次，当我在没有回调的情况下运行训练器时，

i.e. self.trainer = pl.Trainer()

模型运行。