| --- |
| license: mit |
| base_model: runwayml/stable-diffusion-v1-5 |
| tags: |
| - stable-diffusion |
| - text-to-image |
| - diffusers |
| - pytorch |
| - fine-tuned-unet |
| inference: true |
| --- |
| |
| # Stable Diffusion v1-3.0 (Fine-Tuned UNet) |
|
|
| This repository contains fine-tuned **UNet** weights for Stable Diffusion v1.5, trained by [UX4567](https://huggingface.co/UX4567). |
|
|
| ## Model Details |
| - **Base Model:** [runwayml/stable-diffusion-v1-5](https://huggingface.co/runwayml/stable-diffusion-v1-5) |
| - **Model Type:** Text-to-Image Diffusion (UNet component) |
| - **Training Environment:** Google Colab (T4 GPU) |
| - **Precision:** FP16 / 8-bit Adam optimization (`bitsandbytes`) |
|
|
| ## How to Use |
|
|
| Because this repository stores the fine-tuned UNet weights, you can easily load it using the `diffusers` library by injecting it into the base Stable Diffusion pipeline: |
|
|
| ```python |
| import torch |
| from diffusers import StableDiffusionPipeline, UNet2DConditionModel |
| |
| model_id = "UX4567/Stable-Diffusion-v1-3.0" |
| device = "cuda" if torch.cuda.is_available() else "cpu" |
| |
| # 1. Load your fine-tuned UNet in float16 |
| trained_unet = UNet2DConditionModel.from_pretrained( |
| model_id, |
| torch_dtype=torch.float16 |
| ) |
| |
| # 2. Initialize pipeline with runwayml base and your custom UNet |
| pipe = StableDiffusionPipeline.from_pretrained( |
| "runwayml/stable-diffusion-v1-5", |
| unet=trained_unet, |
| torch_dtype=torch.float16 |
| ).to(device) |
| |
| # 3. Generate an image |
| prompt = "a futuristic cyberpunk city at night" |
| image = pipe(prompt, num_inference_steps=30).images[0] |
| image.save("generated_image.png") |
