Update app.py

app.py

@@ -5,6 +5,7 @@ from decord import VideoReader, cpu
 from PIL import Image
 import numpy as np
 import transformers
+import spaces
 from typing import Dict, Optional, Sequence, List
 
 import subprocess
@@ -81,7 +82,7 @@ def preprocess_qwen(sources, tokenizer: transformers.PreTrainedTokenizer, has_im
     targets = torch.tensor(targets, dtype=torch.long)
     return input_ids
 
-
+@spaces.GPU(duration=120)
 def oryx_inference(video, text):
     vr = VideoReader(video, ctx=cpu(0))
     total_frame_num = len(vr)