Return a protobuf of tokenizer info, callable from Py or C++.
(self)
| 1487 | } |
| 1488 | |
| 1489 | def get_tokenizer(self) -> TokenizerParameters: |
| 1490 | """Return a protobuf of tokenizer info, callable from Py or C++.""" |
| 1491 | try: |
| 1492 | tokenizer_type_val = TokenizerType.DESCRIPTOR.values_by_name[self.config.tokenizer_type].number |
| 1493 | return TokenizerParameters( |
| 1494 | path=self.config.tokenizer_path, |
| 1495 | tokenizer_type=tokenizer_type_val, |
| 1496 | access_token=self.config.hf_access_token, |
| 1497 | use_chat_template=self.config.use_chat_template, |
| 1498 | extra_ids=0, |
| 1499 | ) |
| 1500 | except KeyError as _: |
| 1501 | raise KeyError(f"Unsupported tokenizer type: {self.config.tokenizer_type}") from None |
| 1502 | |
| 1503 | def build_tokenizer(self, metadata: TokenizerParameters) -> tokenizer_api.Tokenizer: |
| 1504 | """Return a tokenizer""" |