Run

Provided you already read the configuration page and prepared all the files, you simply need to

1. open a terminal window and go to webui repository folder
2. launch webui in --api mode
3. open a second terminal window and go again to webui repository
4. activate webui venv
5. go to extensions/sd-webui-bayesian-merger folder
6. run python3 bayesian_merger.py

Now just wait and come back later to check results.

Results

In the logs directory you'll find a folder named bbwm-model_a-model_b-RUN_DATE_TIME containing:

• imgs folder: if --save_imgs was used, here all the generated images will be saved/
• log.json: this contains the scores and the weights for all the iterations.
• best.log: this contains the weights for the best merge of the entire run. First line is the base_alpha value, following the weights in a single line.
• bbwm-model_a-model_b.png: a plot showing the evolution of the score across the iterations.
• bbwm-model_a-model_b-unet.png: an image showing the best weights on the UNET architecture