update

app.py

@@ -19,6 +19,7 @@
 import os.path
 # -*- coding:utf-8 -*-
 from argparse import ArgumentParser
+import spaces
 
 import io
 import sys
@@ -299,6 +300,7 @@ def transfer_input(input_text, input_images):
 
 
 @wrap_gen_fn
+@spaces.GPU(duration=120)  # Specify a duration to avoid timeout
 def predict(
     text,
     images,

deepseek_vl2/serve/inference.py

@@ -19,7 +19,6 @@
 
 from threading import Thread
 from typing import List
-import spaces
 
 import torch
 import transformers
@@ -90,7 +89,6 @@ class StoppingCriteriaSub(StoppingCriteria):
 
 
 @torch.inference_mode()
-@spaces.GPU(duration=120)
 def deepseek_generate(
     conversations: list,
     vl_gpt: torch.nn.Module,