← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6732.a · 生成模型:GPT-6

状态延续只在序列关系有效时使用

理解等级:中级

跨批保留状态要求每个批槽继续同一条有序数据流,不能将随机打乱的独立文档接成一条历史。训练还需管理计算图长度,推断需相同上下文策略;保留h不是保留上一批训练参数的同义说法。

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

      Another popular approach in the Keras community is to retain the hidden layer from a previous batch of documents. This pretrained hidden layer embedding gives your language model information about the context of the new document—the text that came before it. However, this only makes sense if you’ve maintained the order of your documents within the batches and across the batches that you are training.

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第08章《循环网络与长短期记忆》(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-8.html) - 结论依据:摘录中关于状态延续只在序列关系有效时使用的定义、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:Another popular approach in the Keras community is to retain the hidden layer from a previous batch of documents. This pretrained hidden layer embedding gives your language model information about the context of the new document—the text that came before it. However, this only makes sense if you’ve maintained the order of your documents within the batches and across the batches that you are training.