Update handler.py

handler.py

@@ -44,7 +44,7 @@ class EndpointHandler():
             latents = latents.to(device="cuda")
             
             with torch.no_grad():
-                latents_dist = self.vae.encode(latents).latent_dist.sample() * vae.config.scaling_factor
+                latents_dist = self.vae.encode(latents).latent_dist.sample() * self.vae.config.scaling_factor
             
             self.smooth_pipe.enable_xformers_memory_efficient_attention()
             out = self.smooth_pipe(prompt, image=latents_dist).images