ComfyUI/comfy_extras/nodes_cfg.py
comfyanonymous 650838fd6f
Experimental CFGNorm node. (#8942)
This is from the new hidream e1 1 model code. Figured it might be useful as a generic cfg trick.
2025-07-17 04:11:07 -04:00

72 lines
2.4 KiB
Python

import torch
# https://github.com/WeichenFan/CFG-Zero-star
def optimized_scale(positive, negative):
positive_flat = positive.reshape(positive.shape[0], -1)
negative_flat = negative.reshape(negative.shape[0], -1)
# Calculate dot production
dot_product = torch.sum(positive_flat * negative_flat, dim=1, keepdim=True)
# Squared norm of uncondition
squared_norm = torch.sum(negative_flat ** 2, dim=1, keepdim=True) + 1e-8
# st_star = v_cond^T * v_uncond / ||v_uncond||^2
st_star = dot_product / squared_norm
return st_star.reshape([positive.shape[0]] + [1] * (positive.ndim - 1))
class CFGZeroStar:
@classmethod
def INPUT_TYPES(s):
return {"required": {"model": ("MODEL",),
}}
RETURN_TYPES = ("MODEL",)
RETURN_NAMES = ("patched_model",)
FUNCTION = "patch"
CATEGORY = "advanced/guidance"
def patch(self, model):
m = model.clone()
def cfg_zero_star(args):
guidance_scale = args['cond_scale']
x = args['input']
cond_p = args['cond_denoised']
uncond_p = args['uncond_denoised']
out = args["denoised"]
alpha = optimized_scale(x - cond_p, x - uncond_p)
return out + uncond_p * (alpha - 1.0) + guidance_scale * uncond_p * (1.0 - alpha)
m.set_model_sampler_post_cfg_function(cfg_zero_star)
return (m, )
class CFGNorm:
@classmethod
def INPUT_TYPES(s):
return {"required": {"model": ("MODEL",),
"strength": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 100.0, "step": 0.01}),
}}
RETURN_TYPES = ("MODEL",)
RETURN_NAMES = ("patched_model",)
FUNCTION = "patch"
CATEGORY = "advanced/guidance"
def patch(self, model, strength):
m = model.clone()
def cfg_norm(args):
cond_p = args['cond_denoised']
pred_text_ = args["denoised"]
norm_full_cond = torch.norm(cond_p, dim=1, keepdim=True)
norm_pred_text = torch.norm(pred_text_, dim=1, keepdim=True)
scale = (norm_full_cond / (norm_pred_text + 1e-8)).clamp(min=0.0, max=1.0)
return pred_text_ * scale * strength
m.set_model_sampler_post_cfg_function(cfg_norm)
return (m, )
NODE_CLASS_MAPPINGS = {
"CFGZeroStar": CFGZeroStar,
"CFGNorm": CFGNorm,
}