fix: set smaller max_new_tokens

app.py

@@ -143,6 +143,8 @@ def initialize_model():
         tokenizer, model, spt = load_model(MODEL_PATH, SPT_CONFIG_PATH, SPT_CHECKPOINT_PATH)
         spt = spt.to(device)
         model = model.to(device)
+        # limit max new tokens to avoid timeouts
+        model.generation_config.max_new_tokens = 4096
         print("Model initialization completed!")
     
     return tokenizer, model, spt, device