Update app.py

app.py

@@ -56,10 +56,9 @@ random_seed = 42
 video_length = 201
 W = 1024
 H = W
-#device = "cuda" if torch.cuda.is_available() else "cpu"
 
 def get_pipe_image_and_video_predictor():
-    device="cuda"
+    device="cpu"
     vae = AutoencoderKLWan.from_pretrained("./model/vae", torch_dtype=torch.float16)
     transformer = Transformer3DModel.from_pretrained("./model/transformer", torch_dtype=torch.float16)
     scheduler = UniPCMultistepScheduler.from_pretrained("./model/scheduler")
@@ -248,6 +247,10 @@ def track_video(n_frames, video_state):
     images = [cv2.resize(img, (W_, H_)) for img in images]
     video_state["origin_images"] = images
     images = np.array(images)
+
+    video_predictor = build_sam2_video_predictor(config, sam2_checkpoint, device="cuda")
+
+    
     inference_state = video_predictor.init_state(images=images/255, device="cuda")
     video_state["inference_state"] = inference_state