SamplerSASolver - ComfyUI Built-in Node Documentation

The SamplerSASolver node creates and configures a custom sampler for diffusion models.

The SamplerSASolver node creates and configures a custom sampler for diffusion models. It uses the “sa_solver” sampling algorithm with a configurable predictor-corrector scheme and stochastic differential equation (SDE) settings, returning a sampler object that can be plugged into a sampling node. The sde_start_percent and sde_end_percent values are converted into sigma values using the connected model’s sampling schedule to define the interval in which the stochastic component is applied.

Inputs

Parameter	Description	Data Type	Required	Range
model	The diffusion model whose sampling schedule is used to build the sampler	MODEL	Yes	-
eta	Controls the step size scaling factor of the SDE solver (default: 1.0)	FLOAT	No	0.0 - 10.0
sde_start_percent	Starting percentage of the sampling process where the stochastic (SDE) component begins; converted to a sigma value using the model’s schedule (default: 0.2)	FLOAT	No	0.0 - 1.0
sde_end_percent	Ending percentage of the sampling process where the stochastic (SDE) component stops; converted to a sigma value using the model’s schedule (default: 0.8)	FLOAT	No	0.0 - 1.0
s_noise	Controls the amount of noise added during sampling (default: 1.0)	FLOAT	No	0.0 - 100.0
predictor_order	The order of the predictor component in the solver (default: 3)	INT	No	1 - 6
corrector_order	The order of the corrector component in the solver (default: 4)	INT	No	0 - 6
use_pece	Enables the PECE (Predict-Evaluate-Correct-Evaluate) method (default: disabled)	BOOLEAN	No	-
simple_order_2	Enables simplified second-order calculations (default: disabled)	BOOLEAN	No	-

All optional inputs are marked as advanced in the interface.

Outputs

Output Name	Description	Data Type
sampler	A configured sampler object (using the sa_solver algorithm) that can be used by sampling nodes	SAMPLER

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

---

Source fingerprint (SHA-256): 31da2d436665bf533c28b32248f632edab8f6d92372402904702ae954230f98d