The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Add a node to zero out the cond to advanced/conditioning
comfyanonymous2d880fec3a72
diff --git a/nodes.py b/nodes.pyindex 3c0960093..f10515f89 100644--- a/nodes.py+++ b/nodes.py@@ -148,6 +148,25 @@ class ConditioningSetMask:c.append(n)return (c, )+class ConditioningZeroOut:+ @classmethod+ def INPUT_TYPES(s):+ return {"required": {"conditioning": ("CONDITIONING", )}}+ RETURN_TYPES = ("CONDITIONING",)+ FUNCTION = "zero_out"++ CATEGORY = "advanced/conditioning"++ def zero_out(self, conditioning):+ c = []+ for t in conditioning:+ d = t[1].copy()+ if "pooled_output" in d:+ d["pooled_output"] = torch.zeros_like(d["pooled_output"])+ n = [torch.zeros_like(t[0]), d]+ c.append(n)+ return (c, )+class VAEDecode:@classmethoddef INPUT_TYPES(s):@@ -1350,6 +1369,8 @@ NODE_CLASS_MAPPINGS = {"LoadLatent": LoadLatent,"SaveLatent": SaveLatent,++ "ConditioningZeroOut": ConditioningZeroOut,}NODE_DISPLAY_NAME_MAPPINGS = {