This repository provides a suite of tools for inferencing and training Stable Diffusion v1.5 models, developed as part of the NTUT AC03519 project "Generative AI Technology and Applications."

This repository is a reference for education only. It is not actively supported.

This repo includes the following components:

1. Diffusion Model
   A customized Gradio WebUI for running Stable Diffusion v1.5 on Google Colab or a local machine.

2. Textual Inversion
   A simple WebUI for training Stable Diffusion v1.5 Textual Inversion models on Google Colab or a local machine.

3. Dreambooth
   A user-friendly WebUI for training Stable Diffusion v1.5 Dreambooth models on Google Colab.
   Note: This module requires significant VRAM. For best performance, it is recommended to run Dreambooth on a local machine.

4. LoRA
   A streamlined WebUI for training Stable Diffusion v1.5 LoRA models on Google Colab or a local machine. The generated <pytorch_lora_weights_kohya.safetensors> file is compatible with the A1111 WebUI LoRA model.

• Git
• Python 3.10+
• Pytorch on CUDA 12.4
• Google Colab (if you choose to run notebooks in the cloud)
• Local machine setup (if running locally, ensure you have sufficient VRAM for Dreambooth)

1. Clone the repository:

   git clone https://github.com/yourusername/MrFuGenerativeAI.git
   cd MrFuGenerativeAI
   

2. Setup the Environment: In each Jupyter notebook, run the 1st cell to install required python packages.

3. Running the WebUI: For Google Colab, open the corresponding notebook github address and follow the in-notebook instructions.

For local setups, run the provided scripts in VSCode with Jupyter plugin installed to launch the Gradio WebUI.

1. Diffusion Model: Run <MrFu_Stable_Diffusion_WebUI.ipynb> to lauch Gradio WebUI and start inferencing using Stable Diffusion v1.5.

2. Textual Inversion: Run <MrFu_Textual_Inversion.ipynb> to train your custom Textual Inversion model via the provided WebUI.

3. Dreambooth: Run <MrFu_Dreambooth.ipynb>to initiate Dreambooth training using the Colab notebook or local setup, keeping in mind the VRAM requirements.

4. LoRA: Run <MrFu_LoRA.ipynb> to train a LoRA model and generate <pytorch_lora_weights_kohya.safetensors> compatible with A1111 WebUI.

For more detailed instructions, please refer to the individual module documentation within this repository.

Contributions are welcome! If you would like to improve the repository or fix any issues, please fork the repository, make your changes, and submit a pull request.

This project is licensed under the MIT License.

• NTUT AC03519 "Generative AI Technology and Applications" – for the support and inspiration behind this project.

• The Huggingface and Stable Diffusion staff and community for their contributions and insights.