CKPT to Diffusers Conversion

This script is used to convert a checkpoint model to a Diffusers model.

Start by cloning the repo and installing the requirements

git clone https://github.com/danbrown/ckpt-to-diffusers.git 
cd ckpt-to-diffusers

chmod +x ./install.sh

You should have the checkpoint downloaded in a ckpt or safetensors, check the format the original file is using, in my case is a safetensors file. wget -O NAMEHERE.safetensors https://civitai.com/api/download/models/29792

⚠️ ATTENTION: both scripts are for diffusion checkpoint models, not LoRAs.

v1 is only compatible with .ckpt format

v2 is compatible with both .ckpt and .safetensors formats, it's the recommended one

v1

For v1, run the ./v1/app.y script with the checkpoint path and the dump path.

python app.py --checkpoint_path="./NAMEHERE.ckpt" --scheduler_type="lms" --dump_path="./out"

v2

For v2, run the ./v2/convert_diffusers20_original_sd.py script with the checkpoint path, the dump path and the reference model.

Check this file if you need more information: convert_diffusers20_original_sd.md

python ./v2/convert_diffusers20_original_sd.py ./NAMEHERE.safetensors ./out --v1 --reference_model runwayml/stable-diffusion-v1-5

---

Thanks for Kohya and the following links for helping me all this together <3:

• Self-Development JP Safetensors article

• This Kohya notes

• Kohya-ss SD Scripts

• bmaltais Kohya_ss UI

• Diffusers Library

If you have any questions, feel free to open an issue here. Thanks!