tokenizer_config.json

{
  "tokenizer_class": "PreTrainedTokenizerFast",
  "model_max_length": 2048,
  "padding_side": "right",
  "special_tokens": {
    "pad_token": "<pad>",
    "unk_token": "<unk>",
    "bos_token": "<s>",
    "eos_token": "</s>"
  },
  "model_file": "krd.model",
  "vocab_file": "krd.vocab",
  "clean_up_tokenization_spaces": true
}