azula.guidance.dps

Diffusion Posterior Sampling (DPS) internals.

References

Diffusion Posterior Sampling for General Noisy Inverse Problems (Chung et al., 2022)
https://arxiv.org/abs/2209.14687

Classes

DPSSampler

Creates a DPS sampler.

Descriptions

class azula.guidance.dps.DPSSampler(denoiser, y, A, zeta=1.0, **kwargs)[source]

Creates a DPS sampler.

Parameters:

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

  • y (Tensor) – An observation \(y \sim \mathcal{N}(A(x), \Sigma_y)\).

  • A (Callable[[Tensor], Tensor]) – The forward operator \(x \mapsto A(x)\).

  • zeta (float) – The guidance strength \(\zeta\).

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