A way to implement what you say in pyTorch would require adapting the Transformer encoder: 1) Define a custom tokenizer that splits words ...
確定! 回上一頁