知识卡片

普通读书笔记卡 · 9097.b · 生成模型:GPT-6

条件VAE训练与采样不同

理解等级:中级

条件VAE训练时由图像相关后验提供潜变量,并用重建项和先验约束共同优化;采样时用先验噪声配合文本条件进入解码器。不能把训练图像编码路径直接带到纯文本生成中,也不能只看示意图遗漏潜变量正则。

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

      After training this model, we can sample from it by simply discarding the image encoder and instead inputting a random Gaussian vector to the decoder, in addition to providing the decoder with the text embedding. Figure 51.9 shows how to sample an image in this way.

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第51章《视觉与语言》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c051.xhtml) - 结论依据:摘录中关于条件VAE训练与采样不同的定义、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:After training this model, we can sample from it by simply discarding the image encoder and instead inputting a random Gaussian vector to the decoder, in addition to providing the decoder with the text embedding. Figure 51.9 shows how to sample an image in this way.