| import gradio as gr | |
| import torch | |
| from diffusers import DiffusionPipeline | |
| pipeline = DiffusionPipeline.from_pretrained("anton-l/ddpm-butterflies-128").to("cuda") | |
| def diffusion(): | |
| images = [] | |
| for i in range(3): | |
| image = pipeline(num_inference_steps=25).images[0] | |
| images.append(image) | |
| return images | |
| demo = gr.Interface( | |
| fn=diffusion, | |
| inputs=None, | |
| outputs=gr.Gallery(label="generated image", columns=3), | |
| title="Unconditional image generation", | |
| description="An unconditional diffusion model trained on a dataset of butterfly images." | |
| ) | |
| demo.launch(debug=True) |