stablediffusion吧 关注:28,363贴子:98,310
  • 1回复贴,共1

这种报错是显存不足吗?

只看楼主收藏回复

到最后直接一黑,图就没了NansException: A tensor with all NaNs was produced in VAE. This could be because there's not enough precision to represent the picture. Try adding --no-half-vae commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.
Time taken: 28.22sTorch active/reserved: 3559/4724 MiB, Sys VRAM: 8188/8188 MiB (100.0%)
ai绘图这么吃显存吗,笔记本4060只能画大概八九百的分辨率


IP属地:湖南来自Android客户端1楼2023-12-15 23:15回复
    用其他的 vae 试试,如 840000 的。或者有些主模型用自动的就行。


    IP属地:马来西亚来自Android客户端2楼2023-12-16 01:00
    回复