Upload handler.py

handler.py

@@ -158,8 +158,8 @@ class EndpointHandler:
         if IS_PARA: apply_cache_on_pipe(self.pipeline, residual_diff_threshold=0.12)
         gc.collect()
         torch.cuda.empty_cache()
-        self.enable_vae_slicing()
-        self.enable_vae_tiling()
+        self.pipeline.enable_vae_slicing()
+        self.pipeline.enable_vae_tiling()
         if IS_LVRAM:
             self.pipeline.transformer.to("cuda")
             self.pipeline.vae.to("cuda")