This model is trained according to DeBERTa-V2(https://huggingface.co/microsoft/deberta-base) for classical Chinese.