| | |
|---|
| | | from typing import Union |
|---|
| | | import logging |
|---|
| | | import torch |
|---|
| | | from typeguard import check_argument_types |
|---|
| | | |
|---|
| | | from funasr.modules.e2e_asr_common import ErrorCalculator |
|---|
| | | from funasr.modules.nets_utils import th_accuracy |
|---|
| | |
|---|
| | | import random |
|---|
| | | import math |
|---|
| | | |
|---|
| | | |
|---|
| | | class MFCCA(FunASRModel): |
|---|
| | | """ |
|---|
| | | Author: Audio, Speech and Language Processing Group (ASLP@NPU), Northwestern Polytechnical University |
|---|
| | |
|---|
| | | frontend: Optional[AbsFrontend], |
|---|
| | | specaug: Optional[AbsSpecAug], |
|---|
| | | normalize: Optional[AbsNormalize], |
|---|
| | | preencoder: Optional[AbsPreEncoder], |
|---|
| | | encoder: AbsEncoder, |
|---|
| | | decoder: AbsDecoder, |
|---|
| | | ctc: CTC, |
|---|
| | | rnnt_decoder: None, |
|---|
| | | rnnt_decoder: None = None, |
|---|
| | | ctc_weight: float = 0.5, |
|---|
| | | ignore_id: int = -1, |
|---|
| | | lsm_weight: float = 0.0, |
|---|
| | |
|---|
| | | report_wer: bool = True, |
|---|
| | | sym_space: str = "<space>", |
|---|
| | | sym_blank: str = "<blank>", |
|---|
| | | preencoder: Optional[AbsPreEncoder] = None, |
|---|
| | | ): |
|---|
| | | assert check_argument_types() |
|---|
| | | assert 0.0 <= ctc_weight <= 1.0, ctc_weight |
|---|
| | | assert rnnt_decoder is None, "Not implemented" |
|---|
| | | |
|---|
