diff --git a/invokeai/backend/stable_diffusion/diffusers_pipeline.py b/invokeai/backend/stable_diffusion/diffusers_pipeline.py
index 51e7b1ee1df..7245997de94 100644
--- a/invokeai/backend/stable_diffusion/diffusers_pipeline.py
+++ b/invokeai/backend/stable_diffusion/diffusers_pipeline.py
@@ -343,6 +343,8 @@ def __init__(
         self._model_group = FullyLoadedModelGroup(self.unet.device)
         self._model_group.install(*self._submodels)
 
+        self.enable_vae_tiling()
+
     def _adjust_memory_efficient_attention(self, latents: torch.Tensor):
         """
         if xformers is available, use it, otherwise use sliced attention.