Skip to main content
This node is designed to modify the sampling behavior of a model by applying a discrete sampling strategy. It allows for the selection of different sampling methods, such as epsilon, v_prediction, lcm, or x0, and optionally adjusts the model’s noise reduction strategy based on the zero-shot noise ratio (zsnr) setting.

Inputs

ParameterDescriptionData TypePython dtype
modelThe model to which the discrete sampling strategy will be applied. This parameter is crucial as it defines the base model that will undergo modification.MODELtorch.nn.Module
samplingSpecifies the discrete sampling method to be applied to the model. The choice of method affects how the model generates samples, offering different strategies for sampling.COMBO[STRING]str
zsnrA boolean flag that, when enabled, adjusts the model’s noise reduction strategy based on the zero-shot noise ratio. This can influence the quality and characteristics of the generated samples.BOOLEANbool

Outputs

ParameterDescriptionData TypePython dtype
modelThe modified model with the applied discrete sampling strategy. This model is now equipped to generate samples using the specified method and adjustments.MODELtorch.nn.Module
This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub