/making-of-yuru-flat

Primary LanguagePython

Yuru-Flat Lora

The Yuru-Flat Lora is distributed under the Fair AI Public License 1.0-SD. In compliance with the license requirements, the following documentation details the steps involved in the creation of this Lora.

Creation Process

  1. Initial Image: Start with the Original.png file available in this repository.

  2. Image Transformation:

    • Use the img2img to convert Original.png to Temp.png.
    • Perform another img2img to transform Temp.png into Target.png.
    • The parameters used in the img2img processes are included in the metadata of the respective PNG files in this repository.

  3. Model Training:

    • Utilize the sd-webui-traintain plugin.
    • Set the plugin to Difference mode using the Difference_Use2ndPass preset.
    • Specify Original.png as the Original and Target.png as the Target.
    • The output from this step will be test_yurucamp.safetensors.

  4. Lora Pruning:

    • Run the loraedit.py script to prune the Lora.
    • The final output, yuru_flat.safetensors, is the pruned Lora ready for use.