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).

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

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).

raw latents -> [0, 1]

Scale the parameters of a module and return it.

Disables custom attention processors and sets the default attention implementation.

Disables custom attention processors and sets the default attention implementation.

Disables custom attention processors and sets the default attention implementation.

Updates the step count.

Performs a single optimization step. Args: closure (callable, optional): A closure that reevaluates the model and

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

Disables the fused QKV projection if enabled. <Tip warning={true}> This API is 🧪 experimental. </Tip>

Unloads LoRA weights.

[0, 1] -> raw latents

Get the weight of wrapped embedding layer.