知识卡片
源目标分词可独立也可共享
翻译任务可分别训练两侧分词器与嵌入,也可采用共享词表;需与数据及模型设计一致。共享词并不自动构成信息泄漏,独立词表也不保证不输出错误语言;训练、生成与反解必须复用各自约定。
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
To keep things simple, you can tokenize your source and target language texts separately, with specialized tokenizers for each. If you use the Hugging Face tokenizers, they will keep track of all of the special tokens you’ll need for a transformer to work on almost any machine learning task:
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》第09章《可堆叠的Transformer》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-9.html)
- 结论依据:摘录中关于源目标分词可独立也可共享的定义、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:To keep things simple, you can tokenize your source and target language texts separately, with specialized tokenizers for each. If you use the Hugging Face tokenizers, they will keep track of all of the special tokens you’ll need for a transformer to work on almost any machine learning task: