Views
No views yet
1# !pip install diffusers
2from diffusers import DiffusionPipeline
3import torch
4
5device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
6model_id = "eurecom-ds/scoresdeve-ema-multi-dsprites-64"
7
8# load model and scheduler
9pipe = DiffusionPipeline.from_pretrained(model_id, trust_remote_code=True)
10pipe.to(device)
11
12
13# run pipeline in inference (sample random noise and denoise)
14generator = torch.Generator(device=device).manual_seed(46)
15image = pipe(
16 generator=generator,
17 batch_size=1,
18 num_inference_steps=1000
19).images
20
21
22# save image
23image[0].save("sde_ve_generated_image.png")