azula.guidance.repaint

RePaint internals.

References

RePaint: Inpainting using Denoising Diffusion Probabilistic Models (Lugmayr et al., 2022)

https://arxiv.org/abs/2201.09865

Classes

RePaintSampler

Creates a RePaint sampler.

Descriptions

class azula.guidance.repaint.RePaintSampler(denoiser, y, mask, iterations=3, **kwargs)

Creates a RePaint sampler.

Parameters:

• denoiser (Denoiser) – A denoiser \(q_\phi(X \mid X_t)\).

• y (Tensor) – An observation \(y = m \odot x\).

• mask (BoolTensor) – The observation mask \(m\).

• iterations (int) – The number of RePaint iterations per step.

• kwargs – Keyword arguments passed to azula.sample.DDIMSampler.