Fix wrong model_max_length

The model has a context window of 2048 (`n_ctx`). The tokenizer should also support the same length.

tokenizer_config.json

@@ -2,7 +2,7 @@
   "add_prefix_space": false,
   "bos_token": "<|endoftext|>",
   "eos_token": "<|endoftext|>",
-  "model_max_length": 1024,
+  "model_max_length": 2048,
   "name_or_path": "gpt2",
   "special_tokens_map_file": null,
   "tokenizer_class": "GPT2Tokenizer",