Initialize Transformer. Args: specaug: TODO. specaug_conf: Configuration dict for specaug. normalize: TODO. normalize_conf: Configuration dict for normalize. encoder: TODO. encoder_co
(
self,
specaug: str = None,
specaug_conf: dict = None,
normalize: str = None,
normalize_conf: dict = None,
encoder: str = None,
encoder_conf: dict = None,
ctc_conf: dict = None,
input_size: int = 80,
vocab_size: int = -1,
ignore_id: int = -1,
blank_id: int = 0,
sos: int = 1,
eos: int = 2,
length_normalized_loss: bool = False,
**kwargs,
)
| 19 | """CTC-attention hybrid Encoder-Decoder model""" |
| 20 | |
| 21 | def __init__( |
| 22 | self, |
| 23 | specaug: str = None, |
| 24 | specaug_conf: dict = None, |
| 25 | normalize: str = None, |
| 26 | normalize_conf: dict = None, |
| 27 | encoder: str = None, |
| 28 | encoder_conf: dict = None, |
| 29 | ctc_conf: dict = None, |
| 30 | input_size: int = 80, |
| 31 | vocab_size: int = -1, |
| 32 | ignore_id: int = -1, |
| 33 | blank_id: int = 0, |
| 34 | sos: int = 1, |
| 35 | eos: int = 2, |
| 36 | length_normalized_loss: bool = False, |
| 37 | **kwargs, |
| 38 | ): |
| 39 | |
| 40 | """Initialize Transformer. |
| 41 | |
| 42 | Args: |
| 43 | specaug: TODO. |
| 44 | specaug_conf: Configuration dict for specaug. |
| 45 | normalize: TODO. |
| 46 | normalize_conf: Configuration dict for normalize. |
| 47 | encoder: TODO. |
| 48 | encoder_conf: Configuration dict for encoder. |
| 49 | ctc_conf: Configuration dict for ctc. |
| 50 | input_size: Size/dimension parameter. |
| 51 | vocab_size: Size/dimension parameter. |
| 52 | ignore_id: TODO. |
| 53 | blank_id: TODO. |
| 54 | sos: TODO. |
| 55 | eos: TODO. |
| 56 | length_normalized_loss: TODO. |
| 57 | **kwargs: Additional keyword arguments. |
| 58 | """ |
| 59 | super().__init__() |
| 60 | |
| 61 | if specaug is not None: |
| 62 | specaug_class = tables.specaug_classes.get(specaug) |
| 63 | specaug = specaug_class(**specaug_conf) |
| 64 | if normalize is not None: |
| 65 | normalize_class = tables.normalize_classes.get(normalize) |
| 66 | normalize = normalize_class(**normalize_conf) |
| 67 | encoder_class = tables.encoder_classes.get(encoder) |
| 68 | encoder = encoder_class(input_size=input_size, **encoder_conf) |
| 69 | encoder_output_size = encoder.output_size() |
| 70 | |
| 71 | if ctc_conf is None: |
| 72 | ctc_conf = {} |
| 73 | ctc = CTC(odim=vocab_size, encoder_output_size=encoder_output_size, **ctc_conf) |
| 74 | |
| 75 | self.blank_id = blank_id |
| 76 | self.sos = sos if sos is not None else vocab_size - 1 |
| 77 | self.eos = eos if eos is not None else vocab_size - 1 |
| 78 | self.vocab_size = vocab_size |
nothing calls this directly
no test coverage detected