speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge branch 'dev_gzf' of github.com:alibaba-damo-academy/FunASR into dev_gzf

Browse Source 
merge
This commit is contained in:
游雁 2024-02-21 19:23:46 +08:00
commit ee3e3ff570
6 changed files with 37 additions and 47 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

0
funasr/models/rwkv_bat/model.py
View File

17
funasr/models/rwkv_bat/rwkv.py
View File

@ -1,16 +1,15 @@
"""Receptance Weighted Key Value (RWKV) block definition. #!/usr/bin/env python3
# -*- encoding: utf-8 -*-
Based/modified from https://github.com/BlinkDL/RWKV-LM/blob/main/RWKV-v4/src/model.py # Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
# MIT License (https://opensource.org/licenses/MIT)
"""
from typing import Dict, Optional, Tuple
import torch import torch
from typing import Dict, Optional, Tuple
from funasr.models.rwkv_bat.rwkv_attention import EncoderSelfAttention, DecoderSelfAttention
from funasr.models.rwkv_bat.rwkv_feed_forward import FeedForward
from funasr.models.transformer.layer_norm import LayerNorm from funasr.models.transformer.layer_norm import LayerNorm
from funasr.models.rwkv_bat.rwkv_feed_forward import FeedForward
from funasr.models.rwkv_bat.rwkv_attention import EncoderSelfAttention, DecoderSelfAttention
class RWKV(torch.nn.Module): class RWKV(torch.nn.Module):
"""RWKV module. """RWKV module.

15
funasr/models/rwkv_bat/rwkv_attention.py
View File

@ -1,17 +1,14 @@
"""Attention (time mixing) modules for RWKV block. #!/usr/bin/env python3
# -*- encoding: utf-8 -*-
Based/Modified from https://github.com/BlinkDL/RWKV-LM/blob/main/RWKV-v4/src/model.py. # Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
# MIT License (https://opensource.org/licenses/MIT)
Some variables are renamed according to https://github.com/huggingface/transformers/blob/main/src/transformers/models/rwkv/modeling_rwkv.py.
""" # noqa
import math import math
from importlib.util import find_spec import torch
from pathlib import Path from pathlib import Path
from importlib.util import find_spec
from typing import List, Optional, Tuple, Union from typing import List, Optional, Tuple, Union
import torch
wkv_kernel_encoder = None wkv_kernel_encoder = None
wkv_kernel_decoder = None wkv_kernel_decoder = None

18
funasr/models/rwkv_bat/rwkv_encoder.py
View File

@ -1,17 +1,20 @@
"""RWKV encoder definition for Transducer models.""" #!/usr/bin/env python3
# -*- encoding: utf-8 -*-
import math # Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
from typing import Dict, List, Optional, Tuple # MIT License (https://opensource.org/licenses/MIT)
import torch import torch
from typing import Dict, List, Optional, Tuple
from funasr.models.encoder.abs_encoder import AbsEncoder from funasr.register import tables
from funasr.models.rwkv_bat.rwkv import RWKV from funasr.models.rwkv_bat.rwkv import RWKV
from funasr.models.transformer.layer_norm import LayerNorm from funasr.models.transformer.layer_norm import LayerNorm
from funasr.models.rwkv_bat.rwkv_subsampling import RWKVConvInput
from funasr.models.transformer.utils.nets_utils import make_source_mask from funasr.models.transformer.utils.nets_utils import make_source_mask
from funasr.models.rwkv_bat.rwkv_subsampling import RWKVConvInput
class RWKVEncoder(AbsEncoder):
@tables.register("encoder_classes", "RWKVEncoder")
class RWKVEncoder(torch.nn.Module):
"""RWKV encoder module. """RWKV encoder module.
Based on https://arxiv.org/pdf/2305.13048.pdf. Based on https://arxiv.org/pdf/2305.13048.pdf.
@ -44,6 +47,7 @@ class RWKVEncoder(AbsEncoder):
subsampling_factor: int =4, subsampling_factor: int =4,
time_reduction_factor: int = 1, time_reduction_factor: int = 1,
kernel: int = 3, kernel: int = 3,
**kwargs,
) -> None: ) -> None:
"""Construct a RWKVEncoder object.""" """Construct a RWKVEncoder object."""
super().__init__() super().__init__()

14
funasr/models/rwkv_bat/rwkv_feed_forward.py
View File

@ -1,14 +1,10 @@
"""Feed-forward (channel mixing) module for RWKV block. #!/usr/bin/env python3
# -*- encoding: utf-8 -*-
Based/Modified from https://github.com/BlinkDL/RWKV-LM/blob/main/RWKV-v4/src/model.py # Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
# MIT License (https://opensource.org/licenses/MIT)
Some variables are renamed according to https://github.com/huggingface/transformers/blob/main/src/transformers/models/rwkv/modeling_rwkv.py.
""" # noqa
from typing import List, Optional, Tuple
import torch import torch
from typing import List, Optional, Tuple
class FeedForward(torch.nn.Module): class FeedForward(torch.nn.Module):

20
funasr/models/rwkv_bat/rwkv_subsampling.py
View File

@ -1,19 +1,13 @@
#!/usr/bin/env python3 #!/usr/bin/env python3
# -*- coding: utf-8 -*- # -*- encoding: utf-8 -*-
# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
# MIT License (https://opensource.org/licenses/MIT)
# Copyright 2019 Shigeki Karita
# Apache 2.0 (http://www.apache.org/licenses/LICENSE-2.0)
"""Subsampling layer definition."""
import numpy as np
import torch
import torch.nn.functional as F
from funasr.models.transformer.embedding import PositionalEncoding
import logging
from funasr.models.scama.utils import sequence_mask
from funasr.models.transformer.utils.nets_utils import sub_factor_to_params, pad_to_len
from typing import Optional, Tuple, Union
import math import math
import torch
from typing import Optional, Tuple, Union
from funasr.models.transformer.utils.nets_utils import pad_to_len
class TooShortUttError(Exception): class TooShortUttError(Exception):
"""Raised when the utt is too short for subsampling. """Raised when the utt is too short for subsampling.