AI Model Fine-tuning Tool
7.7k 2026-05-01
XavierXiao/Dreambooth-Stable-Diffusion
This project implements Google's Dreambooth technique on Stable Diffusion, enabling users to fine-tune a text-to-image model with a few custom examples for personalized image generation.
Core Features
Dreambooth implementation on Stable Diffusion
Fine-tuning of the entire diffusion model (Unet)
Support for regularization images during training
GPU memory optimization via gradient checkpointing
Customizable learning rates and regularization weights
Quick Start
python main.py --base configs/stable-diffusion/v1-finetune_unfrozen.yaml -t --actual_resume /path/to/original/stable-diffusion/sd-v1-4-full-ema.ckpt -n <job name> --gpus 0, --data_root /root/to/training/images --reg_data_root /root/to/regularization/images --class_word <xxx>Detailed Introduction
Google's Dreambooth technique enables the personalization of text-to-image models using a minimal set of user-provided images. This project ports Dreambooth, originally designed for Imagen, to the more accessible Stable Diffusion model. By fine-tuning the entire diffusion model, it allows users to generate specific subjects or styles from just a few examples, overcoming the limitations of models like Textual Inversion which only optimize word embeddings. It provides a practical way for individuals to create highly customized AI-generated content.