Update tokenizer_config.json

The [NeuralBeagle](https://huggingface.co/mlabonne/NeuralBeagle14-7B) model [recently updated its tokenization config to fix the bugs with it](https://huggingface.co/mlabonne/NeuralBeagle14-7B/commit/7d8fe89cdddfc9c28dbee535f5f53d38b5440d74), and this simply copy-pastes this in here.

tokenizer_config.json

@@ -27,16 +27,23 @@
       "special": true
     }
   },
-  "additional_special_tokens": [],
+  "additional_special_tokens": [
+    "<unk>",
+    "<s>",
+    "</s>"
+  ],
   "bos_token": "<s>",
   "clean_up_tokenization_spaces": false,
   "eos_token": "</s>",
   "legacy": true,
-  "model_max_length": 1000000000000000019884624838656,
-  "pad_token": null,
+  "model_max_length": 8192,
+  "pad_token": "</s>",
+  "padding_side": "left",
   "sp_model_kwargs": {},
   "spaces_between_special_tokens": false,
+  "split_special_tokens": false,
   "tokenizer_class": "LlamaTokenizer",
   "unk_token": "<unk>",
-  "use_default_system_prompt": false
-}
+  "chat_template": "{% for message in messages %}{{bos_token + message['role'] + '\n' + message['content'] + eos_token + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ bos_token + 'assistant\n' }}{% endif %}",
+  "use_default_system_prompt": true
+}