The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Some comments to say what the vram state options mean.
comfyanonymousfed0a4dd2985
diff --git a/comfy/model_management.py b/comfy/model_management.pyindex a492ca6b9..1a8a1be17 100644--- a/comfy/model_management.py+++ b/comfy/model_management.py@@ -4,12 +4,12 @@ from comfy.cli_args import argsimport torchclass VRAMState(Enum):- DISABLED = 0- NO_VRAM = 1+ DISABLED = 0 #No vram present: no need to move models to vram+ NO_VRAM = 1 #Very low vram: enable all the options to save vramLOW_VRAM = 2NORMAL_VRAM = 3HIGH_VRAM = 4- SHARED = 5+ SHARED = 5 #No dedicated vram: memory shared between CPU and GPU but models still need to be moved between both.class CPUState(Enum):GPU = 0