Update app.py

app.py

@@ -160,7 +160,7 @@ def generate(segment, image, prompt, size, guidance_scale, num_inference_steps,
                     torch.cat([image_latents] * 2)
                 )        
                 latent_model_input = torch.cat([latent_model_input, latent_image_input], dim=1)
-                timestep = t.expand(latents.shape[0]).to(latents.dtype)
+                timestep = t.repeat(latent_model_input.shape[0]).to(torch.float32)
                 with torch.no_grad():
                   noise_pred = pipe.transformer(
                     hidden_states=latent_model_input,