/a1111-sd-webui-lycoris

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

Primary LanguagePythonApache License 2.0Apache-2.0

a1111-sd-webui-lycoris

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

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