Skip to content

An extension for stable-diffusion-webui to load lycoris models.

License

Notifications You must be signed in to change notification settings

AI2Hub/a1111-sd-webui-lycoris

 
 

Repository files navigation

a1111-sd-webui-lycoris [sd-webui<1.5.0]

NOTICE

This extension is for stable-diffusion-webui < 1.5.0.

After 1.5.0 version of sd-webui, all the functionality of this extension have been merged into sd-webui built-in lora.

If you have any problems like "lack of some functionalities" "want more compatibilities", you can open issues in this repo or sd-webui but plz don't say "BRING THIS BACK". Everyone can BRING THIS BACK, but not me.

(I mean, it is left for old version user, and merged into sd-webui for new version user. it is not disappeared)

I will continuously maintain this extension, adding new algorithms and fixing bugs. However, for the new version sd-webui, all updates will be included in the built-in loras of sd-webui.


An extension for loading lycoris model in sd-webui. I made this stand alone extension (Use sd-webui's extra networks api) to avoid some conflict with other loras extensions.

How to install

There are some options you can choose to install this extension

  • Open the extension tab and go to "available" tab, search "lycoris" to find this extension and then install it.
  • Open the extension tab and go to "from url" tab, copy-paste the url of this repo (https://github.com/KohakuBlueleaf/a1111-sd-webui-lycoris) and click install.
  • Manually clone this repo to the extension folder or download the zip.

The version of stable-diffusion-webui

Lot of people struggling on some bugs or unexpected behavior after install the extension. We do some research and test on it and can only get this conclusion:
Make sure your stable diffusion webui is after commit: a9fed7c3
(a9fed7c3 itself should work, but if you meet problem on that commit, you should consider to update your sd-webui)

LyCORIS

https://github.com/KohakuBlueleaf/LyCORIS

usage

Install it and restart the webui Don't use "Apply and restart UI", please restart the webui process

And you will find "LyCORIS" tab in the extra networks page Use <lyco:MODEL:WEIGHT> to utilize the lycoris model image

The launch parameter --lyco-dir can be used to define LyCORIS models location path

Launch parameter

aka command_args in sd-webui.

Argument What it does default
--lyco-dir the model path for lycoris models use models/LyCORIS as model directory
--lyco-debug debug mode, print out more information when generating images off
--lyco-patch-lora replace the built-in lora ext with lycoris ext. will use models/lora(or --lora-dir) and format. Arguments system still working with this. This is for who are migrating from locon ext, will let some lora-related ext become unusable off

Arguments

sd-webui use this format to use extra networks: <TYPE:MODEL_NAME:arg1:arg2:arg3...:argn>
With more and more different algorithm be implemented into lycoris, the arguments become more.
So I design this arg system to use it more easily(Maybe):

<lyco:MODEL:arg1:arg2:k1=v1:k2=v2>

For example, we have te/unet/dyn these 3 arguments, if you want te=1, unet=0.5, dyn=13, you can use it like these:
<lyco:Model:1:0.5:13>
<lyco:Model:1:0.5:dyn=13>
<lyco:Model:1:unet=0.5:dyn=13>
And if you specify ALL the key name, you can ignore the order:
(or, actually, we only count the args, no k-v pair, so dyn=13:unet=1:0.5 also work, but 0.5 is for te (the first argument))
<lyco:Model:dyn=13:te=1:unet=0.5>

And since te=1 is default value, you can also do it like this:
<lyco:Model:unet=0.5:dyn=13>

And here is the list for arguments:

Argument What it does default type and value
te the weight for text encoder float: 1.0
unet the weight for UNet, when it is None, it use same value as te float: None
dyn How many row you want to utilize when using dylora, if you set to 0, it will disable the dylora int: None

About

An extension for stable-diffusion-webui to load lycoris models.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%