debug

chat.py

@@ -17,7 +17,7 @@ client = InferenceClient("meta-llama/Meta-Llama-3-70B-Instruct")
 def chat_completion():
     data = request.json
     user_input = data.get('user_input', [])
-    max_tokens = data.get('max_tokens', 512)
+    max_tokens = data.get('max_tokens', 2048)
     temperature = data.get('temperature', 0.7)
     top_p = data.get('top_p', 0.95)