知识卡片

普通读书笔记卡 · 6655.e · 生成模型:GPT-6

基准吞吐应核对单位与样本规模

理解等级:中级

字符数、词元数和每秒词元数不是同一单位,书中60k、160k等还来自不同修订时的文本。需保存实际输入、长度、输出数量和计时方法,再估总量;不能将一个章节耗时直接无条件换成处理一万本书所需天数。

这张卡片看不懂?

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

      SpaCy can parse the AsciiDoc text for a chapter in this book in about 5 seconds. First, download the AsciiDoc text file for this chapter: >>> import requests >>> text = requests.get('https://proai.org/nlpia2-ch2.adoc').text >>> f'{round(len(text) / 10_000)}0k' #1 '60k' #1 Δivides by 10,000 and then rounds so that doctests continue to pass as text is revised There were about 160,000 ASCII characters in the AsciiDoc file

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第02章《思维的词元》(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-2.html) - 结论依据:摘录中关于基准吞吐应核对单位与样本规模的定义、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:SpaCy can parse the AsciiDoc text for a chapter in this book in about 5 seconds. First, download the AsciiDoc text file for this chapter: >>> import requests >>> text = requests.get('https://proai.org/nlpia2-ch2.adoc').text >>> f'{round(len(text) / 10_000)}0k' #1 '60k' #1 Δivides by 10,000 and then rounds so that doctests continue to pass as text is revised There were about 160,000 ASCII characters in the AsciiDoc file