tokenizer_config.json

{
    "pad_token": "<|endoftext|>",
    "unk_token": "<|endoftext|>",
    "bos_token": "<|startoftext|>",
    "eos_token": "<|endoftext|>",
    "do_clean_text": false
}