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.pyindex 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",)