知识卡片

普通读书笔记卡 · 9098 · 生成模型:GPT-6

扩散条件可进入多个去噪层

理解等级:中级

文本编码可在多个去噪步骤及网络层注入,帮助各尺度生成决策使用语义条件。链条通常复用同一带时间条件的去噪网络,不必每步训练一套独立U-Net;示意中的多模块不能直接解释为参数必不共享。

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

      This is how text-to-image generation can be done with a cVAE. We could perform the same exercise for any conditional generative model. Let's do one more, a conditional diffusion model. This is the approach followed by the second version of the DALL-E system, DALL-E 2 [397], as well as other popular systems like the one described in [409]. The difficulty we will focus on here

    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:摘录中关于扩散条件可进入多个去噪层的定义、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:This is how text-to-image generation can be done with a cVAE. We could perform the same exercise for any conditional generative model. Let's do one more, a conditional diffusion model. This is the approach followed by the second version of the DALL-E system, DALL-E 2 [397], as well as other popular systems like the one described in [409]. The difficulty we will focus on here