Merge pull request #956 from alibaba-damo-academy/chenmengzheAAA-patch-4

chenmengzheAAA

2023-09-14 2a66366be4c2715870e4859fd5a5db6e8a9dc00a

 funasr/models/data2vec.py

@@ -10,15 +10,14 @@
from typing import Tuple

import torch
from typeguard import check_argument_types

from funasr.layers.abs_normalize import AbsNormalize
from funasr.models.base_model import FunASRModel
from funasr.models.encoder.abs_encoder import AbsEncoder
from funasr.models.frontend.abs_frontend import AbsFrontend
from funasr.models.preencoder.abs_preencoder import AbsPreEncoder
from funasr.models.specaug.abs_specaug import AbsSpecAug
from funasr.torch_utils.device_funcs import force_gatherable
from funasr.train.abs_espnet_model import AbsESPnetModel

if LooseVersion(torch.__version__) >= LooseVersion("1.6.0"):
    from torch.cuda.amp import autocast
@@ -29,7 +28,7 @@
        yield


class Data2VecPretrainModel(AbsESPnetModel):
class Data2VecPretrainModel(FunASRModel):
    """Data2Vec Pretrain model"""

    def __init__(
@@ -37,10 +36,9 @@
            frontend: Optional[AbsFrontend],
            specaug: Optional[AbsSpecAug],
            normalize: Optional[AbsNormalize],
            preencoder: Optional[AbsPreEncoder],
            encoder: AbsEncoder,
            preencoder: Optional[AbsPreEncoder] = None,
    ):
        assert check_argument_types()

        super().__init__()

@@ -57,7 +55,6 @@
            speech_lengths: torch.Tensor,
    ) -> Tuple[torch.Tensor, Dict[str, torch.Tensor], torch.Tensor]:
        """Frontend + Encoder + Calc loss

        Args:
            speech: (Batch, Length, ...)
            speech_lengths: (Batch, )
@@ -106,7 +103,6 @@
            speech_lengths: torch.Tensor,
    ):
        """Frontend + Encoder.

        Args:
            speech: (Batch, Length, ...)
            speech_lengths: (Batch, )