set model_max_length to the maximum length of model context (131072 tokens)

tokenizer_config.json

@@ -9011,7 +9011,7 @@
   "eos_token": "</s>",
   "extra_special_tokens": {},
   "legacy": true,
-  "model_max_length": 1000000000000000019884624838656,
+  "model_max_length": 131072,
   "pad_token": "<pad>",
   "processor_class": "PixtralProcessor",
   "tokenizer_class": "LlamaTokenizerFast",