Skip to content

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.

Make LCM sampler use the model noise scaling function.

comfyanonymous150a3e946fe4
diff --git a/comfy/k_diffusion/sampling.py b/comfy/k_diffusion/sampling.py
index 761c2e0ef..57518c7b6 100644
--- a/comfy/k_diffusion/sampling.py
+++ b/comfy/k_diffusion/sampling.py
@@ -748,7 +748,7 @@ def sample_lcm(model, x, sigmas, extra_args=None, callback=None, disable=None, n
x = denoised
if sigmas[i + 1] > 0:
- x += sigmas[i + 1] * noise_sampler(sigmas[i], sigmas[i + 1])
+ x = model.inner_model.inner_model.model_sampling.noise_scaling(sigmas[i + 1] * s_in, noise_sampler(sigmas[i], sigmas[i + 1]), x)
return x