DDIMSolver#

class deepinv.sampling.DDIMSolver(timesteps=None, t_start=None, t_end=None, num_steps=None, eta=0.0, variance='small', rng=None)[source]#

Bases: AncestralSolver

DDIM solver for reverse-time diffusion SDEs.

Sampler of DDIM [1] with the stochasticity parameter \(\eta\). The default \(\eta = 0\) gives the deterministic DDIM sampler and \(\eta = 1\) gives the DDPM sampler, see deepinv.sampling.DDPMSolver.

The alpha of the SDE is ignored, and replaced on each step of deepinv.sampling.AncestralSolver by

\[\alpha_\eta = \frac{\log\left(1 - \eta^2 (1 - r^2)\right)}{2 \log r}, \quad r = \frac{\sigma(t+dt)}{\sigma(t)}.\]

This relation depends on the step: \(\alpha_\eta = \eta\) for \(\eta \in \{0, 1\}\), and \(\alpha_\eta\) tends to \(\eta^2\) for small steps.

With variance="large", the noise is scaled by \(s(t) \sigma(t) / (s(t+dt) \sigma(t+dt))\), see deepinv.sampling.AncestralSolver. For \(\eta = 1\), this is the DDPM sampler with the variance \(\beta_t\) of the forward transition.

Parameters:

  • timesteps (torch.Tensor, numpy.ndarray, list) – time steps at which the SDE will be discretized.

  • t_start (float) – the starting time of the SDE, optional. If not provided, it will be inferred from the timesteps argument.

  • t_end (float) – the ending time of the SDE, optional. If not provided, it will be inferred from the timesteps argument.

  • num_steps (int) – the number of time steps for the SDE, optional. If not provided, it will be inferred from the timesteps argument.

  • eta (float) – the stochasticity parameter \(\eta \in [0, 1]\) of DDIM. Default to 0.

  • variance (str) – the variance of the noise added on each step, either "small" (default) or "large", see deepinv.sampling.AncestralSolver.

  • rng (torch.Generator) – A random number generator for reproducibility.

References: