This repository contains tools and resources to assist in the fine-tuning of text-to-image diffusion models with Kohya-ss. It includes a Jupyter Notebook for organizing JSON files and large images, which helps track and quantify the fine-tuning process. Additionally, the repository contains a ComfyUI workflow for XY testing of fine-tuned models, and a Python script to create grids.
Fine-Tuner used:
- Load and process Kohya SS JSON files, display/customize important settings and XY Grids.
- Create images in a grid format with folder names and prompts.
- Organize and visualize fine-tuning progress.
- ComfyUI workflow for XY testing of fine-tuned models and LoRA.
- Python script to make grids
Make sure to install the necessary dependencies listed in requirements.txt.
-
Clone the repository:
git clone https://github.com/yourusername/Quant-T2I-Diffusion.git cd Quant-T2I-Diffusion -
Install the required dependencies:
pip install -r requirements.txt
-
Run the Jupyter Notebook:
Open the Jupyter Notebook and add JSON file path to a code cell.
-
Create Image Grid:
Use the provided script or ComfyUI workflow to create a grid image from your custom model. Add the image path to the notebook code.
The repository also includes a ComfyUI workflow for XY testing of the fine-tuned models. Detailed instructions for setting up and using the ComfyUI workflow can be found in the comfyui-workflow/ directory.
Feel free to fork this repository and submit pull requests. For major changes, please open an issue first to discuss what you would like to change.
This project is licensed under the MIT License - see the LICENSE file for details.