Update app.py

app.py

@@ -157,7 +157,7 @@ def generate(segment, image, prompt, size, guidance_scale, num_inference_steps,
                 latent_model_input = torch.cat([latent_model_input, latent_image_input], dim=1)
                 timestep = t.expand(latents.shape[0]).to(latents.dtype)
                 with torch.no_grad():
-                  noise_pred = self.transformer(
+                  noise_pred = pipe.transformer(
                     hidden_states=latent_model_input,
                     timestep=timestep,
                     encoder_hidden_states=prompt_embeds,
@@ -173,9 +173,9 @@ def generate(segment, image, prompt, size, guidance_scale, num_inference_steps,
                 latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0]
     intermediate_latents_cpu = latents.detach().cpu()
     if segment==8:
-        latents = latents.to(self.vae.dtype) / self.vae.config.scaling_factor
-        video = self.vae.decode(latents, return_dict=False)[0]
-        video = self.video_processor.postprocess_video(video, output_type=output_type)
+        latents = latents.to(self.vae.dtype) / pipe.vae.config.scaling_factor
+        video = pipe.vae.decode(latents, return_dict=False)[0]
+        video = pipe.video_processor.postprocess_video(video, output_type=output_type)
         # return HunyuanVideoPipelineOutput(frames=video)
         save_dir = f"./"
         video_out_file = f"{save_dir}/{seed}.mp4"