Update app.py

app.py

@@ -49,8 +49,7 @@ def prompt_model(model_id, system_prompt, user_prompt, schema):
 
 def download_model(base_model_id):
     tokenizer = AutoTokenizer.from_pretrained(base_model_id)
-    model = AutoModelForCausalLM.from_pretrained(base_model_id)
-    
+
     ###
     rope_scaling = {
         "type": "linear",
@@ -60,6 +59,7 @@ def download_model(base_model_id):
     model.config.rope_scaling = rope_scaling
     ###
     
+    model = AutoModelForCausalLM.from_pretrained(base_model_id)    
     model.save_pretrained(base_model_id)
     return tokenizer