SDEOutput

class deepinv.sampling.SDEOutput(sample, trajectory, timesteps, nfe)

Bases: dict

A container for storing the output of an SDE solver, that behaves like a dict but allows access with the attribute syntax.

Attributes: :attr torch.Tensor sample: the final samples of the sampling process, of shape (B, C, H, W). :attr torch.Tensor trajectory: the trajectory of the sampling process, of shape (num_steps, B, C, H, W) if full_trajectory is True, otherwise of shape (B, C, H, W). :attr torch.Tensor timesteps: the time steps at which the samples were taken, of shape (num_steps,). :attr int nfe: the number of function evaluations performed during the integration.