BigrBear | Stylized「LoCon」
Training models take alot of time and effort! Support me by leaving Feed backs and Reviews! ☕
This is LyCORIS Network as such you will need an extension to utilize them.
(I recommend using it natively instead of loading the network via KohyaSS Extension as the Kohya extension likes to break at times.)
My models are made to be used at [ Weight 1 ] Unless it is being combined with other Networks.
There are 2 Versions to this model.
「Normal」and「Hard」
「Normal」is very flexible and you're able to prompt just about anything with it.
「Hard」is Fixated to the dataset and will create backgrounds very similar to the dataset with the artists stone like texture on every background image. Its not recommended to use but its there for the sake of being there.
TOS! Terms of Service!
You are not allowed to sell generated images
You are not allowed to host this model on generation services
Exceptions : Unless the style of this model is merged enough with different styles making it not easily distinguishable then the TOS does not apply to such derivative models.
Here's a google colab to run this LoCon along with other models If you don't meet the requirements to run Stable Diffusion Locally.
Description
Weight Decay 0.15
1e-3 Unet
7e-5 TE
Linear Dim 16
Linear Alpha 8
Conv Dim 16
Conv Alpha8
bf16
FAQ
Comments (4)
Nice job ! What parameter did you use ? I never manage to get a result this clean...
Kohya Additional Networks tab will let u view every parameter. If I post the link in the comments you wont be able to see it untill the moderators review it.
but just get the Hard version and additional networks TAB u can view the parameters
(NORMAL) is just a dynamic resize with sv_fro 0.85 at 16Dim
I said just get the (hard) version because I don't think the dynamic resized version prints out the training metadata. but for prompting just use the (Normal) version as its superior and more flexable
@Hosioka ok thanks