Update app.py

app.py

@@ -61,7 +61,7 @@ def past_kv_to_device(past_kv, device):
     return tuple((k.to(device).detach(), v.to(device).detach()) for k, v in past_kv)
 
 def detach_past_kv(past_kv):
-    return tuple((k.detach().numpy().tolist(), v.detach().numpy().tolist()) for k, v in past_kv)
+    return tuple((k.cpu().detach().numpy().tolist(), v.cpu().detach().numpy().tolist()) for k, v in past_kv)
 
 @spaces.GPU
 def set_past_key_values():