tokenizer_config.json

{
  "vocab_size": 32000,
  "model_max_length": 2048,
  "tokenizer_class": "PreTrainedTokenizerFast",
  "bos_token": "<bos>",
  "eos_token": "<eos>",
  "pad_token": "<pad>",
  "unk_token": "<unk>",
  "additional_special_tokens": [
    "<think>",
    "</think>",
    "<step>",
    "</step>",
    "<answer>",
    "</answer>",
    "<equation>",
    "</equation>",
    "<op>",
    "</op>",
    "<lang>",
    "</lang>"
  ],
  "clean_up_tokenization_spaces": false
}