知识卡片
生成预算按词元而非单词控制
max_new_tokens限制新增token数量,max_length在本章接口中包含输入长度。词元可能是词的一部分或标点,不等于完整单词;预算还应考虑上下文容量及停止符,不能将生成5词元写成最多5个词。
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
#1 Sets the random seed to force an LLM to generate consistent answers from one run to the next #2 Tells the LLM to generate no more than 5 words #3 Generates 10 guesses (possible answers) And when ChatGPT launched, the GPT-3 model wasn’t any better at common-sense reasoning. As the model was scaled up in size and complexity, it was able to memorize more
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》第10章《真实世界中的大语言模型》(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-10.html)
- 结论依据:摘录中关于生成预算按词元而非单词控制的定义、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:#1 Sets the random seed to force an LLM to generate consistent answers from one run to the next #2 Tells the LLM to generate no more than 5 words #3 Generates 10 guesses (possible answers) And when ChatGPT launched, the GPT-3 model wasn’t any better at common-sense reasoning. As the model was scaled up in size and complexity, it was able to memorize more