tokenization_makemore.py

from transformers import PreTrainedTokenizer

AUTO_MAP = {
    "AutoTokenizer": "tokenization_makemore.MakemoreTokenizer"
}


class MakemoreTokenizer(PreTrainedTokenizer):
    def __init__(self, **kwargs):
        self._stoi = {'.': 0, **{chr(ord('a') + i): i + 1 for i in range(26)}}
        self._itos = {v: k for k, v in self._stoi.items()}
        super().__init__(**kwargs)

    @property
    def vocab_size(self):
        return 27

    def get_vocab(self):
        return dict(self._stoi)

    def _tokenize(self, text):
        return list(text.lower())

    def _convert_token_to_id(self, token):
        return self._stoi.get(token, 0)

    def _convert_id_to_token(self, index):
        return self._itos.get(index, '.')

    def save_vocabulary(self, save_directory, filename_prefix=None):
        return ()