知识卡片
序列上限与向量维度分开考虑
序列上限限制模型实际看到多少token,向量维度决定输出长度和存储。长文本截断会丢失尾部,即使输出维度很大也不能恢复未读内容;应以具体配置检查,不能沿用书中某供应商的历史参数作通用值。
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
- · 表示学习与嵌入
原书出处 · 1 处
- · 查看原书依据
sequence length of this particular model is 512, which means the input text is restricted to 512 tokens, beyond which it will be truncated. The embedding vector is made up of floating-point numbers, which by themselves are not interpretable. We will discuss techniques for interpreting embeddings later in this chapter.
参考来源
- 位置:《Designing Large Language Model Applications》第11章《表示学习与嵌入》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch11.html)
- 结论依据:摘录中关于序列上限与向量维度分开考虑的定义、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:sequence length of this particular model is 512, which means the input text is restricted to 512 tokens, beyond which it will be truncated. The embedding vector is made up of floating-point numbers, which by themselves are not interpretable. We will discuss techniques for interpreting embeddings later in this chapter.