Update app.py

app.py

@@ -12,7 +12,7 @@ def generate_response(prompt, model_provider, temperature, top_p, max_tokens, re
             messages=[{"role": "user", "content": prompt}],
             temperature=temperature,
             top_p=top_p,
-            max_tokens=max_tokens,
+            max_tokens=int(max_tokens),
             presence_penalty=repetition_penalty,
             stream=False
         )