yuezhengrong/sd15-lora-artifacts

SDv1.5 Artifacts-500 LoRA Usage Guide

Introduction

SDv1.5 Artifacts-500 LoRA is a fine-tuned model based on Stable Diffusion v1.5, specifically optimized for generating patterns from the artifacts-500 dataset(artifacts-500.zip). Using LoRA (Low-Rank Adaptation) technology, the model has been adapted to produce higher-quality patterns relevant to the dataset.

Usage Instructions

1. Download Stable Diffusion v1.5 Weights

Before you begin, ensure you have downloaded the pre-trained weights for Stable Diffusion v1.5. You can download the weights from the official Stable Diffusion repository.

2. Prepare LoRA Weights

We have trained LoRA weights for the Artifacts-500 dataset. You can download the trained LoRA weights from the following links:

• LoRA weights after 100 epochs: artifacts_100epoch_lora.safetensors

3. Test the Model

After downloading the weights, you can use the generate.py script to test the model's performance. Follow these steps:

Install Dependencies

Ensure you have the following Python libraries installed:

pip install diffusers transformers torch

4. Run the Test Script

To test the model with the LoRA weights trained for 1 epoch:

python generate.py

The param lcm_speedup decide use lcm speed up or not.

View the Results

The generated images will be saved to the specified paths:

Results after 100 epochs: 100epoch_test_results.png

Here are the example results:

•