Skip to main content
The SamplerLCM node provides an LCM (Latent Consistency Model) sampler with tunable per-step noise parameters. It allows you to control the noise applied at each sampling step, enabling fine-grained adjustment of the sampling process.

Inputs

ParameterDescriptionData TypeRequiredRange
s_noisePer-step noise multiplier at the first step. A value of 1.0 matches the model’s training noise scale. (default: 1.0)FLOATYes0.0 to 64.0 (step: 0.01)
s_noise_endPer-step noise multiplier at the last step. Set equal to s_noise for a constant noise schedule. (default: 1.0)FLOATYes0.0 to 64.0 (step: 0.01)
noise_clip_stdClamps the per-step noise to within +/- N standard deviations. A value of 0 disables clamping. (default: 0.0)FLOATYes0.0 to 10.0 (step: 0.01)

Outputs

Output NameDescriptionData Type
SAMPLERThe configured LCM sampler object, ready to be used in a sampling workflow.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): e6f9007f66625baeee8850018784187cf45117591c443f117c593eef547ada98