The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
pop clip vision keys after loading them.
comfyanonymouscd930d4e7fb5
diff --git a/comfy/clip_vision.py b/comfy/clip_vision.pyindex 7a59ef6e2..2036175b8 100644--- a/comfy/clip_vision.py+++ b/comfy/clip_vision.py@@ -21,7 +21,7 @@ class ClipVisionModel():size=224)def load_sd(self, sd):- self.model.load_state_dict(sd, strict=False)+ return self.model.load_state_dict(sd, strict=False)def encode_image(self, image):img = torch.clip((255. * image[0]), 0, 255).round().int()@@ -59,7 +59,13 @@ def load_clipvision_from_sd(sd):else:json_config = os.path.join(os.path.dirname(os.path.realpath(__file__)), "clip_vision_config_vitl.json")clip = ClipVisionModel(json_config)- clip.load_sd(sd)+ m, u = clip.load_sd(sd)+ u = set(u)+ keys = list(sd.keys())+ for k in keys:+ if k not in u:+ t = sd.pop(k)+ del treturn clipdef load(ckpt_path):