知识卡片
评估样例必须记录真实模型身份
代码指定的模型ID、实际配置与分词器,才决定实验对象。本章首次pipeline使用openai-gpt,却在说明中称GPT-2;后面gpt2又是另一模型。对比能力或规模前必须确认版本,不能把不同模型的输出当同一基线。
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 see what’s really going on, it can help to test an early version of the LLM behind ChatGPT. Unfortunately, the only OpenAI LLM that you can download is GPT-2, released in 2019. All these years later, they still have not released the full-size, 1.5-billion parameter model, but instead, they have released a half-size model with 775 million parameters. Nonetheless, clever open source developers were
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:摘录中关于评估样例必须记录真实模型身份的定义、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 see what’s really going on, it can help to test an early version of the LLM behind ChatGPT. Unfortunately, the only OpenAI LLM that you can download is GPT-2, released in 2019. All these years later, they still have not released the full-size, 1.5-billion parameter model, but instead, they have released a half-size model with 775 million parameters. Nonetheless, clever open source developers were