Update app.py

app.py

@@ -112,7 +112,9 @@ def load_and_prepare_model():
     #pipe.enable_vae_tiling()
     pipe.to(device=device, dtype=torch.bfloat16)
     pipe.unet.set_attn_processor(AttnProcessor2_0())
-    pipe.vae = vaeXL #.to('cpu') #.to(torch.bfloat16)
+    #pipe.vae = vaeXL #.to('cpu') #.to(torch.bfloat16)
+    pipe.vae = vaeXL.to(device=device, dtype=torch.bfloat16) # for non-offload
+    
     pipe.vae.set_default_attn_processor()
     return pipe