Add repetition_penalty and max_new_tokens to prevent loops

generation_config.json

@@ -9,5 +9,7 @@
   "temperature": 0.7,
   "top_k": 20,
   "top_p": 0.8,
+  "repetition_penalty": 1.15,
+  "max_new_tokens": 2048,
   "transformers_version": "4.57.3"
-}
+}