[新手包] 窮撚玩 AI - 用 stable diffusion 砌本寫真畀自己 J

係，我有 10 個 google account ...

因為有時懶 (同埋 web UI 虧虧地)，所以唔係好想用佢個 PNG Info

日韓台妹同chilloutmix 都好重要
要成日backup
而且成日條link 都唔同左
起VM要注意

樓主可唔可以講下點解我down咗啲新嘅lora
upload咗上個colab之後都係唔work？

我係用隔離post講嗰個Backward Virgin Killer Sweater
https://civitai.com/models/10867/backwards-virgin-killer-sweater

之後我再裝其他lora都好似唔work

但我打咗prompt上去之後都係冇出到嘢

同埋我發現用Korean Doll係特別容易變異
用Japanese Doll反而最穩陣

有無人知colab pay as you go $77.8 可以玩幾耐？

樓主我用你 v4個notebook但係都仲係唔得，我係咪有嘢set錯？

chilloutmix 好似出番嚟
https://civitai.com/models/6424/chilloutmix

我都有呢個問題, 以我有限既理解能力似係
File "/content/stable-diffusion-webui/modules/devices.py", line 152, in test_for_nans
呢行出問題, 旦係我唔識點停個process 或者撚入去check 條code, 所以都係要等樓主或其他高手.

https://civitai.com/models/6424/chilloutmix

This model has been republished and its ownership transferred to Civitai with the full permissions of the model creator. They have asked that all identifying information about them be removed from the model details.

https://twitter.com/tasuku2023/status/1630751390791536640?s=46&t=S6Jfxfk3ZnqmBRM8OjXsMA

如果我部電腦有情感
呢幾晚睇到我啲 prompt 應該崩潰咗
感謝樓主

師兄你係咪冇入任何 prompt 就 generate? 我見個 error 係話個AI 唔知你想要乜

我剩係打咗girl係咪簡單得濟

去civitai到事旦搵張圖，抄佢個prompt gen住先

聖人 🫡

around 80 個鐘，所以一唔用就記得 disconnect and delete 你個instance

師兄 .. 對自己有返 d 要求，唔好係女都啱

型

強貼留名

背景一有路人就好假

我上網搵咗啲prompt試吓gen, 都係唔得
都係彈咗依句
NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.

1070 ti整張512x512大約要30幾秒
遲D諗住換張3060 12gb，就可以幾秒整副圖同自己train個dreambooth出黎

即係新chilloutmix仲多啲野？