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.

Fix some potential issues.

comfyanonymous430a8334c500
diff --git a/comfy/clip_vision.py b/comfy/clip_vision.py
index 1206c680d..e085186ef 100644
--- a/comfy/clip_vision.py
+++ b/comfy/clip_vision.py
@@ -92,8 +92,11 @@ def load_clipvision_from_sd(sd, prefix="", convert_keys=False):
json_config = os.path.join(os.path.dirname(os.path.realpath(__file__)), "clip_vision_config_g.json")
elif "vision_model.encoder.layers.30.layer_norm1.weight" in sd:
json_config = os.path.join(os.path.dirname(os.path.realpath(__file__)), "clip_vision_config_h.json")
- else:
+ elif "vision_model.encoder.layers.22.layer_norm1.weight" in sd:
json_config = os.path.join(os.path.dirname(os.path.realpath(__file__)), "clip_vision_config_vitl.json")
+ else:
+ return None
+
clip = ClipVisionModel(json_config)
m, u = clip.load_sd(sd)
if len(m) > 0:
diff --git a/comfy/sd.py b/comfy/sd.py
index 48ee5721b..c364b723c 100644
--- a/comfy/sd.py
+++ b/comfy/sd.py
@@ -434,10 +434,11 @@ def load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True, o
if output_clip:
w = WeightsLoader()
clip_target = model_config.clip_target()
- clip = CLIP(clip_target, embedding_directory=embedding_directory)
- w.cond_stage_model = clip.cond_stage_model
- sd = model_config.process_clip_state_dict(sd)
- load_model_weights(w, sd)
+ if clip_target is not None:
+ clip = CLIP(clip_target, embedding_directory=embedding_directory)
+ w.cond_stage_model = clip.cond_stage_model
+ sd = model_config.process_clip_state_dict(sd)
+ load_model_weights(w, sd)
left_over = sd.keys()
if len(left_over) > 0: