Created by: liu0haha123

Paddle 版本1.7.2 在使用ErnieModelForTokenClassification 是需要获取中间步骤的 的这两个参数，能否通过调用父类ERINEModel的forward方法来获取 pooled和encoded 或者其他能满足该要求的方法

    def forward(self, *args, **kwargs):
        """
        Args:
            labels (optional, `Variable` of shape [batch_size, seq_len]): 
                ground truth label id for each token
        Returns:
            loss (`Variable` of shape []):
                Cross entropy loss mean over batch and time, ignore positions where label == -100
                if labels not set, returns None
            logits (`Variable` of shape [batch_size, seq_len, hidden_size]):
                output logits of classifier
        """

        labels = kwargs.pop('labels', None)
        pooled, encoded = super(ErnieModelForTokenClassification, self).forward(*args, **kwargs)