You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
PaddleSpeech/deepspeech/modules/encoder_layer.py

285 lines
11 KiB

E2E/Streaming Transformer/Conformer ASR (#578) * add cmvn and label smoothing loss layer * add layer for transformer * add glu and conformer conv * add torch compatiable hack, mask funcs * not hack size since it exists * add test; attention * add attention, common utils, hack paddle * add audio utils * conformer batch padding mask bug fix #223 * fix typo, python infer fix rnn mem opt name error and batchnorm1d, will be available at 2.0.2 * fix ci * fix ci * add encoder * refactor egs * add decoder * refactor ctc, add ctc align, refactor ckpt, add warmup lr scheduler, cmvn utils * refactor docs * add fix * fix readme * fix bugs, refactor collator, add pad_sequence, fix ckpt bugs * fix docstring * refactor data feed order * add u2 model * refactor cmvn, test * add utils * add u2 config * fix bugs * fix bugs * fix autograd maybe has problem when using inplace operation * refactor data, build vocab; add format data * fix text featurizer * refactor build vocab * add fbank, refactor feature of speech * refactor audio feat * refactor data preprare * refactor data * model init from config * add u2 bins * flake8 * can train * fix bugs, add coverage, add scripts * test can run * fix data * speed perturb with sox * add spec aug * fix for train * fix train logitc * fix logger * log valid loss, time dataset process * using np for speed perturb, remove some debug log of grad clip * fix logger * fix build vocab * fix logger name * using module logger as default * fix * fix install * reorder imports * fix board logger * fix logger * kaldi fbank and mfcc * fix cmvn and print prarams * fix add_eos_sos and cmvn * fix cmvn compute * fix logger and cmvn * fix subsampling, label smoothing loss, remove useless * add notebook test * fix log * fix tb logger * multi gpu valid * fix log * fix log * fix config * fix compute cmvn, need paddle 2.1 * add cmvn notebook * fix layer tools * fix compute cmvn * add rtf * fix decoding * fix layer tools * fix log, add avg script * more avg and test info * fix dataset pickle problem; using 2.1 paddle; num_workers can > 0; ckpt save in exp dir;fix setup.sh; * add vimrc * refactor tiny script, add transformer and stream conf * spm demo; librisppech scripts and confs * fix log * add librispeech scripts * refactor data pipe; fix conf; fix u2 default params * fix bugs * refactor aishell scripts * fix test * fix cmvn * fix s0 scripts * fix ds2 scripts and bugs * fix dev & test dataset filter * fix dataset filter * filter dev * fix ckpt path * filter test, since librispeech will cause OOM, but all test wer will be worse, since mismatch train with test * add comment * add syllable doc * fix ds2 configs * add doc * add pypinyin tools * fix decoder using blank_id=0 * mmseg with pybind11 * format code
3 years ago
# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Encoder self-attention layer definition."""
from typing import Optional
from typing import Tuple
import paddle
from paddle import nn
from deepspeech.utils.log import Log
logger = Log(__name__).getlog()
__all__ = ["TransformerEncoderLayer", "ConformerEncoderLayer"]
class TransformerEncoderLayer(nn.Layer):
"""Encoder layer module."""
def __init__(
self,
size: int,
self_attn: nn.Layer,
feed_forward: nn.Layer,
dropout_rate: float,
normalize_before: bool=True,
concat_after: bool=False, ):
"""Construct an EncoderLayer object.
Args:
size (int): Input dimension.
self_attn (nn.Layer): Self-attention module instance.
`MultiHeadedAttention` or `RelPositionMultiHeadedAttention`
instance can be used as the argument.
feed_forward (nn.Layer): Feed-forward module instance.
`PositionwiseFeedForward`, instance can be used as the argument.
dropout_rate (float): Dropout rate.
normalize_before (bool):
True: use layer_norm before each sub-block.
False: to use layer_norm after each sub-block.
concat_after (bool): Whether to concat attention layer's input and
output.
True: x -> x + linear(concat(x, att(x)))
False: x -> x + att(x)
"""
super().__init__()
self.self_attn = self_attn
self.feed_forward = feed_forward
self.norm1 = nn.LayerNorm(size, epsilon=1e-12)
self.norm2 = nn.LayerNorm(size, epsilon=1e-12)
self.dropout = nn.Dropout(dropout_rate)
self.size = size
self.normalize_before = normalize_before
self.concat_after = concat_after
# concat_linear may be not used in forward fuction,
# but will be saved in the *.pt
self.concat_linear = nn.Linear(size + size, size)
def forward(
self,
x: paddle.Tensor,
mask: paddle.Tensor,
pos_emb: Optional[paddle.Tensor]=None,
E2E/Streaming Transformer/Conformer ASR (#578) * add cmvn and label smoothing loss layer * add layer for transformer * add glu and conformer conv * add torch compatiable hack, mask funcs * not hack size since it exists * add test; attention * add attention, common utils, hack paddle * add audio utils * conformer batch padding mask bug fix #223 * fix typo, python infer fix rnn mem opt name error and batchnorm1d, will be available at 2.0.2 * fix ci * fix ci * add encoder * refactor egs * add decoder * refactor ctc, add ctc align, refactor ckpt, add warmup lr scheduler, cmvn utils * refactor docs * add fix * fix readme * fix bugs, refactor collator, add pad_sequence, fix ckpt bugs * fix docstring * refactor data feed order * add u2 model * refactor cmvn, test * add utils * add u2 config * fix bugs * fix bugs * fix autograd maybe has problem when using inplace operation * refactor data, build vocab; add format data * fix text featurizer * refactor build vocab * add fbank, refactor feature of speech * refactor audio feat * refactor data preprare * refactor data * model init from config * add u2 bins * flake8 * can train * fix bugs, add coverage, add scripts * test can run * fix data * speed perturb with sox * add spec aug * fix for train * fix train logitc * fix logger * log valid loss, time dataset process * using np for speed perturb, remove some debug log of grad clip * fix logger * fix build vocab * fix logger name * using module logger as default * fix * fix install * reorder imports * fix board logger * fix logger * kaldi fbank and mfcc * fix cmvn and print prarams * fix add_eos_sos and cmvn * fix cmvn compute * fix logger and cmvn * fix subsampling, label smoothing loss, remove useless * add notebook test * fix log * fix tb logger * multi gpu valid * fix log * fix log * fix config * fix compute cmvn, need paddle 2.1 * add cmvn notebook * fix layer tools * fix compute cmvn * add rtf * fix decoding * fix layer tools * fix log, add avg script * more avg and test info * fix dataset pickle problem; using 2.1 paddle; num_workers can > 0; ckpt save in exp dir;fix setup.sh; * add vimrc * refactor tiny script, add transformer and stream conf * spm demo; librisppech scripts and confs * fix log * add librispeech scripts * refactor data pipe; fix conf; fix u2 default params * fix bugs * refactor aishell scripts * fix test * fix cmvn * fix s0 scripts * fix ds2 scripts and bugs * fix dev & test dataset filter * fix dataset filter * filter dev * fix ckpt path * filter test, since librispeech will cause OOM, but all test wer will be worse, since mismatch train with test * add comment * add syllable doc * fix ds2 configs * add doc * add pypinyin tools * fix decoder using blank_id=0 * mmseg with pybind11 * format code
3 years ago
mask_pad: Optional[paddle.Tensor]=None,
output_cache: Optional[paddle.Tensor]=None,
cnn_cache: Optional[paddle.Tensor]=None,
) -> Tuple[paddle.Tensor, paddle.Tensor, paddle.Tensor]:
"""Compute encoded features.
Args:
x (paddle.Tensor): Input tensor (#batch, time, size).
mask (paddle.Tensor): Mask tensor for the input (#batch, time).
pos_emb (paddle.Tensor): just for interface compatibility
to ConformerEncoderLayer
mask_pad (paddle.Tensor): not used here, it's for interface
compatibility to ConformerEncoderLayer
E2E/Streaming Transformer/Conformer ASR (#578) * add cmvn and label smoothing loss layer * add layer for transformer * add glu and conformer conv * add torch compatiable hack, mask funcs * not hack size since it exists * add test; attention * add attention, common utils, hack paddle * add audio utils * conformer batch padding mask bug fix #223 * fix typo, python infer fix rnn mem opt name error and batchnorm1d, will be available at 2.0.2 * fix ci * fix ci * add encoder * refactor egs * add decoder * refactor ctc, add ctc align, refactor ckpt, add warmup lr scheduler, cmvn utils * refactor docs * add fix * fix readme * fix bugs, refactor collator, add pad_sequence, fix ckpt bugs * fix docstring * refactor data feed order * add u2 model * refactor cmvn, test * add utils * add u2 config * fix bugs * fix bugs * fix autograd maybe has problem when using inplace operation * refactor data, build vocab; add format data * fix text featurizer * refactor build vocab * add fbank, refactor feature of speech * refactor audio feat * refactor data preprare * refactor data * model init from config * add u2 bins * flake8 * can train * fix bugs, add coverage, add scripts * test can run * fix data * speed perturb with sox * add spec aug * fix for train * fix train logitc * fix logger * log valid loss, time dataset process * using np for speed perturb, remove some debug log of grad clip * fix logger * fix build vocab * fix logger name * using module logger as default * fix * fix install * reorder imports * fix board logger * fix logger * kaldi fbank and mfcc * fix cmvn and print prarams * fix add_eos_sos and cmvn * fix cmvn compute * fix logger and cmvn * fix subsampling, label smoothing loss, remove useless * add notebook test * fix log * fix tb logger * multi gpu valid * fix log * fix log * fix config * fix compute cmvn, need paddle 2.1 * add cmvn notebook * fix layer tools * fix compute cmvn * add rtf * fix decoding * fix layer tools * fix log, add avg script * more avg and test info * fix dataset pickle problem; using 2.1 paddle; num_workers can > 0; ckpt save in exp dir;fix setup.sh; * add vimrc * refactor tiny script, add transformer and stream conf * spm demo; librisppech scripts and confs * fix log * add librispeech scripts * refactor data pipe; fix conf; fix u2 default params * fix bugs * refactor aishell scripts * fix test * fix cmvn * fix s0 scripts * fix ds2 scripts and bugs * fix dev & test dataset filter * fix dataset filter * filter dev * fix ckpt path * filter test, since librispeech will cause OOM, but all test wer will be worse, since mismatch train with test * add comment * add syllable doc * fix ds2 configs * add doc * add pypinyin tools * fix decoder using blank_id=0 * mmseg with pybind11 * format code
3 years ago
output_cache (paddle.Tensor): Cache tensor of the output
(#batch, time2, size), time2 < time in x.
cnn_cache (paddle.Tensor): not used here, it's for interface
compatibility to ConformerEncoderLayer
Returns:
paddle.Tensor: Output tensor (#batch, time, size).
paddle.Tensor: Mask tensor (#batch, time).
paddle.Tensor: Fake cnn cache tensor for api compatibility with Conformer (#batch, channels, time').
"""
residual = x
if self.normalize_before:
x = self.norm1(x)
if output_cache is None:
x_q = x
else:
assert output_cache.shape[0] == x.shape[0]
assert output_cache.shape[1] < x.shape[1]
assert output_cache.shape[2] == self.size
chunk = x.shape[1] - output_cache.shape[1]
x_q = x[:, -chunk:, :]
residual = residual[:, -chunk:, :]
mask = mask[:, -chunk:, :]
if self.concat_after:
x_concat = paddle.concat(
(x, self.self_attn(x_q, x, x, mask)), axis=-1)
x = residual + self.concat_linear(x_concat)
else:
x = residual + self.dropout(self.self_attn(x_q, x, x, mask))
if not self.normalize_before:
x = self.norm1(x)
residual = x
if self.normalize_before:
x = self.norm2(x)
x = residual + self.dropout(self.feed_forward(x))
if not self.normalize_before:
x = self.norm2(x)
if output_cache is not None:
x = paddle.concat([output_cache, x], axis=1)
fake_cnn_cache = paddle.zeros([1], dtype=x.dtype)
return x, mask, fake_cnn_cache
class ConformerEncoderLayer(nn.Layer):
"""Encoder layer module."""
def __init__(
self,
size: int,
self_attn: nn.Layer,
feed_forward: Optional[nn.Layer]=None,
feed_forward_macaron: Optional[nn.Layer]=None,
conv_module: Optional[nn.Layer]=None,
dropout_rate: float=0.1,
normalize_before: bool=True,
concat_after: bool=False, ):
"""Construct an EncoderLayer object.
Args:
size (int): Input dimension.
self_attn (nn.Layer): Self-attention module instance.
`MultiHeadedAttention` or `RelPositionMultiHeadedAttention`
instance can be used as the argument.
feed_forward (nn.Layer): Feed-forward module instance.
`PositionwiseFeedForward` instance can be used as the argument.
feed_forward_macaron (nn.Layer): Additional feed-forward module
instance.
`PositionwiseFeedForward` instance can be used as the argument.
conv_module (nn.Layer): Convolution module instance.
`ConvlutionModule` instance can be used as the argument.
dropout_rate (float): Dropout rate.
normalize_before (bool):
True: use layer_norm before each sub-block.
False: use layer_norm after each sub-block.
concat_after (bool): Whether to concat attention layer's input and
output.
True: x -> x + linear(concat(x, att(x)))
False: x -> x + att(x)
"""
super().__init__()
self.self_attn = self_attn
self.feed_forward = feed_forward
self.feed_forward_macaron = feed_forward_macaron
self.conv_module = conv_module
self.norm_ff = nn.LayerNorm(size, epsilon=1e-12) # for the FNN module
self.norm_mha = nn.LayerNorm(size, epsilon=1e-12) # for the MHA module
if feed_forward_macaron is not None:
self.norm_ff_macaron = nn.LayerNorm(size, epsilon=1e-12)
self.ff_scale = 0.5
else:
self.ff_scale = 1.0
if self.conv_module is not None:
self.norm_conv = nn.LayerNorm(
size, epsilon=1e-12) # for the CNN module
self.norm_final = nn.LayerNorm(
size, epsilon=1e-12) # for the final output of the block
self.dropout = nn.Dropout(dropout_rate)
self.size = size
self.normalize_before = normalize_before
self.concat_after = concat_after
self.concat_linear = nn.Linear(size + size, size)
def forward(
self,
x: paddle.Tensor,
mask: paddle.Tensor,
pos_emb: paddle.Tensor,
mask_pad: Optional[paddle.Tensor]=None,
output_cache: Optional[paddle.Tensor]=None,
cnn_cache: Optional[paddle.Tensor]=None,
) -> Tuple[paddle.Tensor, paddle.Tensor, paddle.Tensor]:
"""Compute encoded features.
Args:
x (paddle.Tensor): (#batch, time, size)
mask (paddle.Tensor): Mask tensor for the input (#batch, timetime).
pos_emb (paddle.Tensor): positional encoding, must not be None
for ConformerEncoderLayer.
mask_pad (paddle.Tensor): batch padding mask used for conv module, (B, 1, T).
output_cache (paddle.Tensor): Cache tensor of the encoder output
(#batch, time2, size), time2 < time in x.
cnn_cache (paddle.Tensor): Convolution cache in conformer layer
Returns:
paddle.Tensor: Output tensor (#batch, time, size).
paddle.Tensor: Mask tensor (#batch, time).
paddle.Tensor: New cnn cache tensor (#batch, channels, time').
"""
# whether to use macaron style FFN
if self.feed_forward_macaron is not None:
residual = x
if self.normalize_before:
x = self.norm_ff_macaron(x)
x = residual + self.ff_scale * self.dropout(
self.feed_forward_macaron(x))
if not self.normalize_before:
x = self.norm_ff_macaron(x)
# multi-headed self-attention module
residual = x
if self.normalize_before:
x = self.norm_mha(x)
if output_cache is None:
x_q = x
else:
assert output_cache.shape[0] == x.shape[0]
assert output_cache.shape[1] < x.shape[1]
assert output_cache.shape[2] == self.size
chunk = x.shape[1] - output_cache.shape[1]
x_q = x[:, -chunk:, :]
residual = residual[:, -chunk:, :]
mask = mask[:, -chunk:, :]
x_att = self.self_attn(x_q, x, x, pos_emb, mask)
if self.concat_after:
x_concat = paddle.concat((x, x_att), axis=-1)
x = residual + self.concat_linear(x_concat)
else:
x = residual + self.dropout(x_att)
if not self.normalize_before:
x = self.norm_mha(x)
# convolution module
# Fake new cnn cache here, and then change it in conv_module
new_cnn_cache = paddle.zeros([1], dtype=x.dtype)
if self.conv_module is not None:
residual = x
if self.normalize_before:
x = self.norm_conv(x)
x, new_cnn_cache = self.conv_module(x, mask_pad, cnn_cache)
x = residual + self.dropout(x)
if not self.normalize_before:
x = self.norm_conv(x)
# feed forward module
residual = x
if self.normalize_before:
x = self.norm_ff(x)
x = residual + self.ff_scale * self.dropout(self.feed_forward(x))
if not self.normalize_before:
x = self.norm_ff(x)
if self.conv_module is not None:
x = self.norm_final(x)
if output_cache is not None:
x = paddle.concat([output_cache, x], axis=1)
return x, mask, new_cnn_cache