test.py

from pathlib import Path

from diffusers import DDPMPipeline, UNet2DModel, DDPMScheduler
from diffusers.utils import make_image_grid
import torch


if __name__ == '__main__':

    unet=UNet2DModel.from_pretrained(
        'jmemon/ddpm-paintings-128-finetuned-celebahq'
    )

    scheduler = DDPMScheduler.from_pretrained(
        'jmemon/ddpm-paintings-128-finetuned-celebahq'
    )

    pipeline = DDPMPipeline(unet=unet, scheduler=scheduler).to('mps')
    pipeline.enable_attention_slicing()

    images = pipeline(batch_size=4, generator=torch.manual_seed(0), num_inference_steps=50).images
    grid = make_image_grid(images, 2, 2)
    grid.save(Path(__file__).parent / 'grid.png')