| | |
| | | # except TooShortUttError as e: |
| | | # logging.warning(f"Utterance {keys} {e}") |
| | | # hyp = Hypothesis(score=0.0, scores={}, states={}, yseq=[]) |
| | | # results = [[" ", ["<space>"], [2], hyp]] * nbest |
| | | # results = [[" ", ["sil"], [2], hyp]] * nbest |
| | | # |
| | | # # Only supporting batch_size==1 |
| | | # key = keys[0] |
| | |
| | | ngram_weight: float = 0.9, |
| | | nbest: int = 1, |
| | | num_workers: int = 1, |
| | | param_dict: dict = None, |
| | | **kwargs, |
| | | ): |
| | | assert check_argument_types() |
| | |
| | | def _forward(data_path_and_name_and_type, |
| | | raw_inputs: Union[np.ndarray, torch.Tensor] = None, |
| | | output_dir_v2: Optional[str] = None, |
| | | param_dict: dict = None, |
| | | ): |
| | | # 3. Build data-iterator |
| | | if data_path_and_name_and_type is None and raw_inputs is not None: |
| | |
| | | except TooShortUttError as e: |
| | | logging.warning(f"Utterance {keys} {e}") |
| | | hyp = Hypothesis(score=0.0, scores={}, states={}, yseq=[]) |
| | | results = [[" ", ["<space>"], [2], hyp]] * nbest |
| | | results = [[" ", ["sil"], [2], hyp]] * nbest |
| | | |
| | | # Only supporting batch_size==1 |
| | | key = keys[0] |