tokenizer_config.json

{
  "clean_up_tokenization_spaces": true,
  "cls_token": "<cls>",
  "mask_token": "<mask>",
  "model_input_names": [
    "input_ids",
    "attention_mask"
  ],
  "model_max_length": 512,
  "pad_token": "<pad>",
  "sep_token": "<sep>",
  "tokenizer_class": "PreTrainedTokenizerFast"
}