The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
refactor: rename Mahiro CFG to Similarity-Adaptive Guidance (#12172)
Christian Byrned159142615e0
diff --git a/comfy_extras/nodes_mahiro.py b/comfy_extras/nodes_mahiro.pyindex 6459ca8c1..a25226e6d 100644--- a/comfy_extras/nodes_mahiro.py+++ b/comfy_extras/nodes_mahiro.py@@ -10,7 +10,7 @@ class Mahiro(io.ComfyNode):def define_schema(cls):return io.Schema(node_id="Mahiro",- display_name="Mahiro CFG",+ display_name="Positive-Biased Guidance",category="_for_testing",description="Modify the guidance to scale more on the 'direction' of the positive prompt rather than the difference between the negative prompt.",inputs=[@@ -20,27 +20,35 @@ class Mahiro(io.ComfyNode):io.Model.Output(display_name="patched_model"),],is_experimental=True,+ search_aliases=[+ "mahiro",+ "mahiro cfg",+ "similarity-adaptive guidance",+ "positive-biased cfg",+ ],)@classmethoddef execute(cls, model) -> io.NodeOutput:m = model.clone()+def mahiro_normd(args):- scale: float = args['cond_scale']- cond_p: torch.Tensor = args['cond_denoised']- uncond_p: torch.Tensor = args['uncond_denoised']- #naive leap+ scale: float = args["cond_scale"]+ cond_p: torch.Tensor = args["cond_denoised"]+ uncond_p: torch.Tensor = args["uncond_denoised"]+ # naive leapleap = cond_p * scale- #sim with uncond leap+ # sim with uncond leapu_leap = uncond_p * scalecfg = args["denoised"]merge = (leap + cfg) / 2normu = torch.sqrt(u_leap.abs()) * u_leap.sign()normm = torch.sqrt(merge.abs()) * merge.sign()sim = F.cosine_similarity(normu, normm).mean()- simsc = 2 * (sim+1)- wm = (simsc*cfg + (4-simsc)*leap) / 4+ simsc = 2 * (sim + 1)+ wm = (simsc * cfg + (4 - simsc) * leap) / 4return wm+m.set_model_sampler_post_cfg_function(mahiro_normd)return io.NodeOutput(m)