← 返回卡片

原书出处

Designing Large Language Model Applications

First Edition · 核验版本的摘录快照 · 导入于 2026-11-14T13:54:13.932Z

Pre-Tokenization Before we run the tokenizer on the text, we can optionally perform a pre-tokenization step. As mentioned earlier, most tokenizers today employ subword tokenization. A common step is to first perform word tokenization and then feed the output of it to the subword tokenization algorithm. This step is called pre-tokenization. Pre-tokenization is a relatively easy step in English compared to many other languages, since

此预览只包含定位所需的摘录和上下文。

原书定位详情

来源:a19c7933-4eda-4176-aebd-30fe7e1ad292

正文位置:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch03.html

EPUB SHA-256:1b4d2347fec2f949fcddfbaf4fad597a1ae20887a096cfae64c59a619b4781bf