A whole new architecture based on separable convolution is introduced, which can achieve a drastic reduction in trainable parameters on text ...
確定! 回上一頁