diffusers/ddpm_dummy

Dummy diffusion model following architecture of https://github.com/lucidrains/denoising-diffusion-pytorch

Run the model as follows:

from diffusers import UNetModel, GaussianDiffusion
import torch

# 1. Load model
unet = UNetModel.from_pretrained("fusing/ddpm_dummy")

# 2. Do one denoising step with model
batch_size, num_channels, height, width = 1, 3, 32, 32
dummy_noise = torch.ones((batch_size, num_channels, height, width))
time_step = torch.tensor([10])
image = unet(dummy_noise, time_step)

# 3. Load sampler
sampler = GaussianDiffusion.from_config("fusing/ddpm_dummy")

# 4. Sample image from sampler passing the model
image = sampler.sample(model, batch_size=1)

print(image)