updated model_max_length from to 1000000000000000019884624838656 to 32768

the same bug is in the original tokenizer config by

@mistral-ai

tokenizer_config.json

@@ -9011,7 +9011,7 @@
   "eos_token": "</s>",
   "extra_special_tokens": {},
   "legacy": true,
-  "model_max_length": 1000000000000000019884624838656,
+  "model_max_length": 32768,
   "pad_token": "<pad>",
   "padding_side": "left",
   "tokenizer_class": "LlamaTokenizer",