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.

Make the device an optional parameter in the clip loaders.

comfyanonymousc8a3492c22ad
diff --git a/nodes.py b/nodes.py
index 82b1575b7..1b796ced7 100644
--- a/nodes.py
+++ b/nodes.py
@@ -913,6 +913,8 @@ class CLIPLoader:
def INPUT_TYPES(s):
return {"required": { "clip_name": (folder_paths.get_filename_list("text_encoders"), ),
"type": (["stable_diffusion", "stable_cascade", "sd3", "stable_audio", "mochi", "ltxv", "pixart"], ),
+ },
+ "optional": {
"device": (["default", "cpu"], {"advanced": True}),
}}
RETURN_TYPES = ("CLIP",)
@@ -952,6 +954,8 @@ class DualCLIPLoader:
return {"required": { "clip_name1": (folder_paths.get_filename_list("text_encoders"), ),
"clip_name2": (folder_paths.get_filename_list("text_encoders"), ),
"type": (["sdxl", "sd3", "flux", "hunyuan_video"], ),
+ },
+ "optional": {
"device": (["default", "cpu"], {"advanced": True}),
}}
RETURN_TYPES = ("CLIP",)