Fix weights not being copied for lowvram patch

This can probably be improved since the weights themselves never need to be copied as they cannot be changed anyway.
2024-08-28 20:23:08 +02:00
parent 3d9869240b
commit abe5c6a6c9
1 changed files with 1 additions and 1 deletions
@@ -120,7 +120,7 @@ class GGUFModelPatcher(comfy.model_patcher.ModelPatcher):
            if device_to is not None:
                out_weight = weight.to(device_to, copy=True)
            else:
-                out_weight = weight.clone()
+                out_weight = weight.to(weight.device, copy=True) # make sure patches are copied

            if self.patch_on_device:
                patches = move_patch_to_cuda(patches, self.load_device)