tokenizer_config.json

{
    "tokenizer_class": "PreTrainedTokenizerFast",
    "model_input_names": [
        "input_ids"
    ],
    "special_tokens": {
        "bos_token": "<s>",
        "eos_token": "</s>",
        "unk_token": "<unk>",
        "pad_token": "<pad>"
    },
    "vocab_size": 20000,
    "tokenizer_file": "tokenizer.model"
}