知识卡片
模型原料分为四类决策
预训练数据决定接触哪些内容,词表和分词决定如何表示输入,学习目标决定优化什么,架构决定信息如何计算与交互。应用失败可能来自其中不同环节,不能把所有差异都归为参数规模或提示好坏。
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
- · 预训练数据
原书出处 · 1 处
- · 查看原书依据
Ingredients of an LLM Let’s start with the ingredients that go into making an LLM. Broadly speaking, we have: Pre-training data: What’s it trained on? The old computer science adage “garbage in, garbage out” is still accurate when it comes to language modeling. In this chapter we will explore popular pre-training datasets and dig into the various preprocessing steps taken to ensure high-quality data is
参考来源
- 位置:《Designing Large Language Model Applications》第02章《预训练数据》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch02.html)
- 结论依据:摘录中关于模型原料分为四类决策的定义、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:Ingredients of an LLM Let’s start with the ingredients that go into making an LLM. Broadly speaking, we have: Pre-training data: What’s it trained on? The old computer science adage “garbage in, garbage out” is still accurate when it comes to language modeling. In this chapter we will explore popular pre-training datasets and dig into the various preprocessing steps taken to ensure high-quality data is