ModelSamplingContinuousEDM - ComfyUI Built-in Node Documentation

This node is designed to enhance a model’s sampling capabilities by integrating continuous EDM (Energy-based Diffusion Models) sampling techniques. It allows for the dynamic adjustment of the noise levels within the model’s sampling process, offering a more refined control over the generation quality and diversity.

Inputs

Parameter	Data Type	Python dtype	Description
`model`	`MODEL`	`torch.nn.Module`	The model to be enhanced with continuous EDM sampling capabilities. It serves as the foundation for applying the advanced sampling techniques.
`sampling`	COMBO[STRING]	`str`	Specifies the type of sampling to be applied, either ‘eps’ for epsilon sampling or ‘v_prediction’ for velocity prediction, influencing the model’s behavior during the sampling process.
`sigma_max`	`FLOAT`	`float`	The maximum sigma value for noise level, allowing for upper bound control in the noise injection process during sampling.
`sigma_min`	`FLOAT`	`float`	The minimum sigma value for noise level, setting the lower limit for noise injection, thus affecting the model’s sampling precision.

Outputs

Parameter	Data Type	Python dtype	Description
`model`	MODEL	`torch.nn.Module`	The enhanced model with integrated continuous EDM sampling capabilities, ready for further use in generation tasks.

​Inputs

​Outputs

Inputs

Outputs