v11.0. In this work, the emoji prediction model has smaller architecture than BERT base model to save computation power. We denote Transformer.
確定！ 回上一頁
