Home Creators Posts Import Register Favorites Logout

Content

Here's a modified notebook, built to harness the pwoer of custon openai models

https://github.com/Sxela/DiscoDiffusion-Warp/blob/main/Disco_Diffusion_v5_2_Warp_custom_model.ipynb

Should work just fine for LSUN BEDROOM openai model settings, but you can change the settings for your custom ones as well.

So, basically, you should do the following:

Diffusion and CLIP model settings:
 Models Settings

   diffusion_model: custom
   use_secondary_model: off

  Custom model
   
custom_path: (specify path to your model here)

Custom model settings
 
Modify in accordance with your training settings
 (the defaults work for the 1st model from my training colab)

Settings
 
Basic settings
   width_height: [256,256]
(use the size you've trained your model on)


For example, if you've trained or downloaded a custom mode (for example, the faces model from here), you need to put it on your google drive, and then add the full path to it in the colab above under 256x256_openai_comics_faces_v2 by_alex_spirin_114k.pt Custom model -> custom_path

The full path would look like this: /content/drive/MyDrive/somefolder/256x256_openai_comics_faces_v2 by_alex_spirin_114k.pt

where "somefolder" is the folder on your drive where you've put the model.


Check that Models Settings -> diffusion_model: custom 

is set and then press ctrl+f9 to run all cells. Mount your drive when asked to, and wait for your first results. After that you can start changing the settings!

Comments

Dalibor Cee

Please is it possible to add this to vision of chaos disco diffusion 5-5.4 versions .