tokenizer_config.json

{
  "tokenizer_class": "PreTrainedTokenizerFast",
  "bos_token": "<s>",
  "eos_token": "</s>",
  "unk_token": "<unk>",
  "pad_token": "<pad>",
  "mask_token": "<mask>",
  "model_max_length": 512,
  "clean_up_tokenization_spaces": true,
  "add_prefix_space": true,
  "trim_offsets": true
}