知识卡片
模型大小与大小写版本各有条件
层数、隐藏宽度及头数决定参数规模,cased与uncased又改变分词和输入信息。较大BERT在原研究若干基准表现较好,不保证每个应用都优于小模型;应比较资源、数据及目标任务。
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
Borrowing from the discussion on the original transformer earlier in the chapter, for the BERT configurations, L denotes the number of transformer layers. The hidden size is H, and the number of self-attention heads is A. BERT base has dimensions L = 12, H = 768, and A = 12, for a total of 110 million parameters. BERT large has L = 24, H =
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》第09章《可堆叠的Transformer》(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-9.html)
- 结论依据:摘录中关于模型大小与大小写版本各有条件的定义、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:Borrowing from the discussion on the original transformer earlier in the chapter, for the BERT configurations, L denotes the number of transformer layers. The hidden size is H, and the number of self-attention heads is A. BERT base has dimensions L = 12, H = 768, and A = 12, for a total of 110 million parameters. BERT large has L = 24, H =