This model is a diffusion model for unconditional image generation of black-and-white images of handwritten digits.
The number of cycles of the training loop is 10.
Learning rate is 4e-4.
This model has a good generation effect.
1from diffusers import DDPMPipeline
2
3pipeline = DDPMPipeline.from_pretrained('BackTo2014/mnist-demo3')
4image = pipeline().images[0]
5image
