[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/vt/ - Virtual Youtubers

Search:


View post   

>> No.45054763 [View]
File: 1.58 MB, 1536x1152, 00016-2264170635-best quality, gawr gura, hololive english, from side, (((looking back, grimace, glaring, scowl))), blue eyes, white hair, shark.png [View same] [iqdb] [saucenao] [google]
45054763

>>45054411
Anon a decent chunk of the people prompting here are using colab. Barring any particular LoRA someone is using, or a model merge that isn't on the long ass list of models the one for retards can grab? It's the same.
>>45054554
CLIP skip - "at what layer of the CLIP encoder's output do we stop?" CLIP = takes your input prompt, turns it into tokens that the model can actually use. CLIP skip 1 = stops at the last layer. CLIP skip 2 = stops one layer before that. Beyond that, I don't understand how it works, other than that NAI and some other models were trained with CLIP skip 2, which is why you should prompt with CLIP skip 2.
>>45054554
>What are they going to come to my home and force feed me?
Yes? Complain about it, they post more to piss you off. How new?
>Why do that when you can tell him to just boot up your colab? Ah whatever.
Because he's not a GPU-let like us.
>>44976628
>Samegaki Soup, picrel
I will have you know this was not easy. And Gura's ass/tail are actually Bugs' head, I should've Photoshopped his ears out first so it'd know his ass from his face.

Navigation
View posts[+24][+48][+96]