A sampler inspired by DPM-Solver-2 and Algorithm 2 from Karras et al. (2022).

Ancestral sampling with DPM-Solver second-order steps.

DPM-Solver-12 and 23 (adaptive step size). See https://arxiv.org/abs/2206.00927.

DPM-Solver-Fast (fixed step size). See https://arxiv.org/abs/2206.00927.

DPM-Solver++(2M).

DPM-Solver++(2M) SDE.

DPM-Solver++(2M) SDE.

Ancestral sampling with DPM-Solver++(2S) second-order steps.

DPM-Solver++(3M) SDE.

DPM-Solver++ (stochastic).

Implements Algorithm 2 (Euler steps) from Karras et al. (2022).

Ancestral sampling with Euler method steps.

Implements Algorithm 2 (Heun steps) from Karras et al. (2022).

* \brief Access to the linearized coefficient, write-only. * The format of the indexing depends on whether this * matrix is column major (Colu

STATIC METHODS AND OTHER HELPERS * \brief Sets all entries to zero. * Warning: this operation works in-place and therefore violates the copy-on-wr

STATIC METHODS AND OTHER HELPERS * \brief Sets all entries to zero. * Warning: this operation works in-place and therefore violates the copy-on-writ

Returns the state of the warmup scheduler as a :class:`dict`. It contains an entry for every variable in self.__dict__ which is not t

A context manager that sets whether TF32 is allowed on cuDNN or matmul.

* \brief Converts this cuMat matrix to the corresponding Eigen matrix. * Note that Eigen does not support batched matrices. Hence, this * convers

Converts from a tensor to a PIL image.

\returns the tolerance threshold used by the stopping criteria. * \sa setTolerance() */