知识卡片
分词往返检查输入边界
在尚未接入模型时,先对同一文本执行 encode 再 decode,检查分词配置与输入是否匹配。往返一致只验证该样本的文本转换,不验证模型能力、所有 Unicode 边界或特殊标记安全。特殊标记如何编码仍需单独检查。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
原书出处 · 1 处
- · 查看原书依据
Since we have not loaded the LLM yet (the central component shown in figure 2.5), we will first do a simpler dry run using just the tokenizer. Specifically, we will do a tokenization round-trip, that is, we will encode a text into token IDs and then decode those IDs back into text, as illustrated in figure 2.6.
参考来源
- 位置:《Build a Reasoning Model (From Scratch)》第02章《使用预训练模型生成文本》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Build a Reasoning Model (From Scratch)/_epub-src/OEBPS/Text/chapter-2.html)
- 结论依据:摘录中关于分词往返检查输入边界的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:Since we have not loaded the LLM yet (the central component shown in figure 2.5), we will first do a simpler dry run using just the tokenizer. Specifically, we will do a tokenization round-trip, that is, we will encode a text into token IDs and then decode those IDs back into text, as illustrated in figure 2.6.