← 返回刚才阅读的位置

知识卡片

结构图卡 · 6725 · 生成模型:GPT-6

文本卷积各层沿不同轴工作

理解等级:中级

结构图:

flowchart TD
 A[词元编号 B乘L] --> B[嵌入 B乘L乘E]
 B --> C[换轴 B乘E乘L]
 C --> D[多个核宽分支 各B乘F乘L_out]
 D --> E[激活及有效位置汇聚 各B乘F]
 E --> F[拼接 B乘总通道数]
 F --> G[可选dropout与线性logits]
 G --> H[匹配目标的损失或决策]

滤波器参数用正式子模块容器注册;固定宽度来自汇聚与架构约定,不是卷积原始输出天生定长。

这张卡片看不懂?

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

      7.5.1 Network architecture Following are the processing steps and the corresponding shapes of the tensors for each stage of a CNN NLP pipeline. It turns out one of the trickiest things about building a new CNN is keeping track of the shapes of your tensors. Like the bumps and holes on a stack of LEGO bricks, neural network layers need to output tensors with the

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第07章《卷积网络识别文本模式》(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-7.html) - 结论依据:摘录中关于文本卷积各层沿不同轴工作的定义、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:7.5.1 Network architecture Following are the processing steps and the corresponding shapes of the tensors for each stage of a CNN NLP pipeline. It turns out one of the trickiest things about building a new CNN is keeping track of the shapes of your tensors. Like the bumps and holes on a stack of LEGO bricks, neural network layers need to output tensors with the