# Copyright 2020 Emiru Tsunoo
# Apache 2.0 (http://www.apache.org/licenses/LICENSE-2.0)
"""Encoder definition."""
import math
from typing import Optional, Tuple
import torch
from typeguard import check_argument_types
from espnet2.asr.encoder.abs_encoder import AbsEncoder
from espnet.nets.pytorch_backend.nets_utils import make_pad_mask
from espnet.nets.pytorch_backend.transformer.attention import MultiHeadedAttention
from espnet.nets.pytorch_backend.transformer.contextual_block_encoder_layer import (
ContextualBlockEncoderLayer,
)
from espnet.nets.pytorch_backend.transformer.embedding import StreamPositionalEncoding
from espnet.nets.pytorch_backend.transformer.layer_norm import LayerNorm
from espnet.nets.pytorch_backend.transformer.multi_layer_conv import (
Conv1dLinear,
MultiLayeredConv1d,
)
from espnet.nets.pytorch_backend.transformer.positionwise_feed_forward import (
PositionwiseFeedForward,
)
from espnet.nets.pytorch_backend.transformer.repeat import repeat
from espnet.nets.pytorch_backend.transformer.subsampling_without_posenc import (
Conv2dSubsamplingWOPosEnc,
)
[docs]class ContextualBlockTransformerEncoder(AbsEncoder):
"""Contextual Block Transformer encoder module.
Details in Tsunoo et al. "Transformer ASR with contextual block processing"
(https://arxiv.org/abs/1910.07204)
Args:
input_size: input dim
output_size: dimension of attention
attention_heads: the number of heads of multi head attention
linear_units: the number of units of position-wise feed forward
num_blocks: the number of encoder blocks
dropout_rate: dropout rate
attention_dropout_rate: dropout rate in attention
positional_dropout_rate: dropout rate after adding positional encoding
input_layer: input layer type
pos_enc_class: PositionalEncoding or ScaledPositionalEncoding
normalize_before: whether to use layer_norm before the first block
concat_after: whether to concat attention layer's input and output
if True, additional linear will be applied.
i.e. x -> x + linear(concat(x, att(x)))
if False, no additional linear will be applied.
i.e. x -> x + att(x)
positionwise_layer_type: linear of conv1d
positionwise_conv_kernel_size: kernel size of positionwise conv1d layer
padding_idx: padding_idx for input_layer=embed
block_size: block size for contextual block processing
hop_Size: hop size for block processing
look_ahead: look-ahead size for block_processing
init_average: whether to use average as initial context (otherwise max values)
ctx_pos_enc: whether to use positional encoding to the context vectors
"""
def __init__(
self,
input_size: int,
output_size: int = 256,
attention_heads: int = 4,
linear_units: int = 2048,
num_blocks: int = 6,
dropout_rate: float = 0.1,
positional_dropout_rate: float = 0.1,
attention_dropout_rate: float = 0.0,
input_layer: Optional[str] = "conv2d",
pos_enc_class=StreamPositionalEncoding,
normalize_before: bool = True,
concat_after: bool = False,
positionwise_layer_type: str = "linear",
positionwise_conv_kernel_size: int = 1,
padding_idx: int = -1,
block_size: int = 40,
hop_size: int = 16,
look_ahead: int = 16,
init_average: bool = True,
ctx_pos_enc: bool = True,
):
assert check_argument_types()
super().__init__()
self._output_size = output_size
self.pos_enc = pos_enc_class(output_size, positional_dropout_rate)
if input_layer == "linear":
self.embed = torch.nn.Sequential(
torch.nn.Linear(input_size, output_size),
torch.nn.LayerNorm(output_size),
torch.nn.Dropout(dropout_rate),
torch.nn.ReLU(),
)
self.subsample = 1
elif input_layer == "conv2d":
self.embed = Conv2dSubsamplingWOPosEnc(
input_size, output_size, dropout_rate, kernels=[3, 3], strides=[2, 2]
)
self.subsample = 4
elif input_layer == "conv2d6":
self.embed = Conv2dSubsamplingWOPosEnc(
input_size, output_size, dropout_rate, kernels=[3, 5], strides=[2, 3]
)
self.subsample = 6
elif input_layer == "conv2d8":
self.embed = Conv2dSubsamplingWOPosEnc(
input_size,
output_size,
dropout_rate,
kernels=[3, 3, 3],
strides=[2, 2, 2],
)
self.subsample = 8
elif input_layer == "embed":
self.embed = torch.nn.Sequential(
torch.nn.Embedding(input_size, output_size, padding_idx=padding_idx),
)
self.subsample = 1
elif input_layer is None:
self.embed = None
self.subsample = 1
else:
raise ValueError("unknown input_layer: " + input_layer)
self.normalize_before = normalize_before
if positionwise_layer_type == "linear":
positionwise_layer = PositionwiseFeedForward
positionwise_layer_args = (
output_size,
linear_units,
dropout_rate,
)
elif positionwise_layer_type == "conv1d":
positionwise_layer = MultiLayeredConv1d
positionwise_layer_args = (
output_size,
linear_units,
positionwise_conv_kernel_size,
dropout_rate,
)
elif positionwise_layer_type == "conv1d-linear":
positionwise_layer = Conv1dLinear
positionwise_layer_args = (
output_size,
linear_units,
positionwise_conv_kernel_size,
dropout_rate,
)
else:
raise NotImplementedError("Support only linear or conv1d.")
self.encoders = repeat(
num_blocks,
lambda lnum: ContextualBlockEncoderLayer(
output_size,
MultiHeadedAttention(
attention_heads, output_size, attention_dropout_rate
),
positionwise_layer(*positionwise_layer_args),
dropout_rate,
num_blocks,
normalize_before,
concat_after,
),
)
if self.normalize_before:
self.after_norm = LayerNorm(output_size)
# for block processing
self.block_size = block_size
self.hop_size = hop_size
self.look_ahead = look_ahead
self.init_average = init_average
self.ctx_pos_enc = ctx_pos_enc
[docs] def output_size(self) -> int:
return self._output_size
[docs] def forward(
self,
xs_pad: torch.Tensor,
ilens: torch.Tensor,
prev_states: torch.Tensor = None,
is_final=True,
infer_mode=False,
) -> Tuple[torch.Tensor, torch.Tensor, Optional[torch.Tensor]]:
"""Embed positions in tensor.
Args:
xs_pad: input tensor (B, L, D)
ilens: input length (B)
prev_states: Not to be used now.
infer_mode: whether to be used for inference. This is used to
distinguish between forward_train (train and validate) and
forward_infer (decode).
Returns:
position embedded tensor and mask
"""
if self.training or not infer_mode:
return self.forward_train(xs_pad, ilens, prev_states)
else:
return self.forward_infer(xs_pad, ilens, prev_states, is_final)
[docs] def forward_train(
self,
xs_pad: torch.Tensor,
ilens: torch.Tensor,
prev_states: torch.Tensor = None,
) -> Tuple[torch.Tensor, torch.Tensor, Optional[torch.Tensor]]:
"""Embed positions in tensor.
Args:
xs_pad: input tensor (B, L, D)
ilens: input length (B)
prev_states: Not to be used now.
Returns:
position embedded tensor and mask
"""
masks = (~make_pad_mask(ilens)[:, None, :]).to(xs_pad.device)
if isinstance(self.embed, Conv2dSubsamplingWOPosEnc):
xs_pad, masks = self.embed(xs_pad, masks)
elif self.embed is not None:
xs_pad = self.embed(xs_pad)
# create empty output container
total_frame_num = xs_pad.size(1)
ys_pad = xs_pad.new_zeros(xs_pad.size())
past_size = self.block_size - self.hop_size - self.look_ahead
# block_size could be 0 meaning infinite
# apply usual encoder for short sequence
if self.block_size == 0 or total_frame_num <= self.block_size:
xs_pad, masks, _, _, _, _, _ = self.encoders(
self.pos_enc(xs_pad), masks, False, None, None
)
if self.normalize_before:
xs_pad = self.after_norm(xs_pad)
olens = masks.squeeze(1).sum(1)
return xs_pad, olens, None
# start block processing
cur_hop = 0
block_num = math.ceil(
float(total_frame_num - past_size - self.look_ahead) / float(self.hop_size)
)
bsize = xs_pad.size(0)
addin = xs_pad.new_zeros(
bsize, block_num, xs_pad.size(-1)
) # additional context embedding vecctors
# first step
if self.init_average: # initialize with average value
addin[:, 0, :] = xs_pad.narrow(1, cur_hop, self.block_size).mean(1)
else: # initialize with max value
addin[:, 0, :] = xs_pad.narrow(1, cur_hop, self.block_size).max(1)
cur_hop += self.hop_size
# following steps
while cur_hop + self.block_size < total_frame_num:
if self.init_average: # initialize with average value
addin[:, cur_hop // self.hop_size, :] = xs_pad.narrow(
1, cur_hop, self.block_size
).mean(1)
else: # initialize with max value
addin[:, cur_hop // self.hop_size, :] = xs_pad.narrow(
1, cur_hop, self.block_size
).max(1)
cur_hop += self.hop_size
# last step
if cur_hop < total_frame_num and cur_hop // self.hop_size < block_num:
if self.init_average: # initialize with average value
addin[:, cur_hop // self.hop_size, :] = xs_pad.narrow(
1, cur_hop, total_frame_num - cur_hop
).mean(1)
else: # initialize with max value
addin[:, cur_hop // self.hop_size, :] = xs_pad.narrow(
1, cur_hop, total_frame_num - cur_hop
).max(1)
if self.ctx_pos_enc:
addin = self.pos_enc(addin)
xs_pad = self.pos_enc(xs_pad)
# set up masks
mask_online = xs_pad.new_zeros(
xs_pad.size(0), block_num, self.block_size + 2, self.block_size + 2
)
mask_online.narrow(2, 1, self.block_size + 1).narrow(
3, 0, self.block_size + 1
).fill_(1)
xs_chunk = xs_pad.new_zeros(
bsize, block_num, self.block_size + 2, xs_pad.size(-1)
)
# fill the input
# first step
left_idx = 0
block_idx = 0
xs_chunk[:, block_idx, 1 : self.block_size + 1] = xs_pad.narrow(
-2, left_idx, self.block_size
)
left_idx += self.hop_size
block_idx += 1
# following steps
while left_idx + self.block_size < total_frame_num and block_idx < block_num:
xs_chunk[:, block_idx, 1 : self.block_size + 1] = xs_pad.narrow(
-2, left_idx, self.block_size
)
left_idx += self.hop_size
block_idx += 1
# last steps
last_size = total_frame_num - left_idx
xs_chunk[:, block_idx, 1 : last_size + 1] = xs_pad.narrow(
-2, left_idx, last_size
)
# fill the initial context vector
xs_chunk[:, 0, 0] = addin[:, 0]
xs_chunk[:, 1:, 0] = addin[:, 0 : block_num - 1]
xs_chunk[:, :, self.block_size + 1] = addin
# forward
ys_chunk, mask_online, _, _, _, _, _ = self.encoders(
xs_chunk, mask_online, False, xs_chunk
)
# copy output
# first step
offset = self.block_size - self.look_ahead - self.hop_size + 1
left_idx = 0
block_idx = 0
cur_hop = self.block_size - self.look_ahead
ys_pad[:, left_idx:cur_hop] = ys_chunk[:, block_idx, 1 : cur_hop + 1]
left_idx += self.hop_size
block_idx += 1
# following steps
while left_idx + self.block_size < total_frame_num and block_idx < block_num:
ys_pad[:, cur_hop : cur_hop + self.hop_size] = ys_chunk[
:, block_idx, offset : offset + self.hop_size
]
cur_hop += self.hop_size
left_idx += self.hop_size
block_idx += 1
ys_pad[:, cur_hop:total_frame_num] = ys_chunk[
:, block_idx, offset : last_size + 1, :
]
if self.normalize_before:
ys_pad = self.after_norm(ys_pad)
olens = masks.squeeze(1).sum(1)
return ys_pad, olens, None
[docs] def forward_infer(
self,
xs_pad: torch.Tensor,
ilens: torch.Tensor,
prev_states: torch.Tensor = None,
is_final: bool = True,
) -> Tuple[torch.Tensor, torch.Tensor, Optional[torch.Tensor]]:
"""Embed positions in tensor.
Args:
xs_pad: input tensor (B, L, D)
ilens: input length (B)
prev_states: Not to be used now.
Returns:
position embedded tensor and mask
"""
if prev_states is None:
prev_addin = None
buffer_before_downsampling = None
ilens_buffer = None
buffer_after_downsampling = None
n_processed_blocks = 0
past_encoder_ctx = None
else:
prev_addin = prev_states["prev_addin"]
buffer_before_downsampling = prev_states["buffer_before_downsampling"]
ilens_buffer = prev_states["ilens_buffer"]
buffer_after_downsampling = prev_states["buffer_after_downsampling"]
n_processed_blocks = prev_states["n_processed_blocks"]
past_encoder_ctx = prev_states["past_encoder_ctx"]
bsize = xs_pad.size(0)
assert bsize == 1
if prev_states is not None:
xs_pad = torch.cat([buffer_before_downsampling, xs_pad], dim=1)
ilens += ilens_buffer
if is_final:
buffer_before_downsampling = None
else:
n_samples = xs_pad.size(1) // self.subsample - 1
if n_samples < 2:
next_states = {
"prev_addin": prev_addin,
"buffer_before_downsampling": xs_pad,
"ilens_buffer": ilens,
"buffer_after_downsampling": buffer_after_downsampling,
"n_processed_blocks": n_processed_blocks,
"past_encoder_ctx": past_encoder_ctx,
}
return (
xs_pad.new_zeros(bsize, 0, self._output_size),
xs_pad.new_zeros(bsize),
next_states,
)
n_res_samples = xs_pad.size(1) % self.subsample + self.subsample * 2
buffer_before_downsampling = xs_pad.narrow(
1, xs_pad.size(1) - n_res_samples, n_res_samples
)
xs_pad = xs_pad.narrow(1, 0, n_samples * self.subsample)
ilens_buffer = ilens.new_full(
[1], dtype=torch.long, fill_value=n_res_samples
)
ilens = ilens.new_full(
[1], dtype=torch.long, fill_value=n_samples * self.subsample
)
if isinstance(self.embed, Conv2dSubsamplingWOPosEnc):
xs_pad, _ = self.embed(xs_pad, None)
elif self.embed is not None:
xs_pad = self.embed(xs_pad)
# create empty output container
if buffer_after_downsampling is not None:
xs_pad = torch.cat([buffer_after_downsampling, xs_pad], dim=1)
total_frame_num = xs_pad.size(1)
if is_final:
past_size = self.block_size - self.hop_size - self.look_ahead
block_num = math.ceil(
float(total_frame_num - past_size - self.look_ahead)
/ float(self.hop_size)
)
buffer_after_downsampling = None
else:
if total_frame_num <= self.block_size:
next_states = {
"prev_addin": prev_addin,
"buffer_before_downsampling": buffer_before_downsampling,
"ilens_buffer": ilens_buffer,
"buffer_after_downsampling": xs_pad,
"n_processed_blocks": n_processed_blocks,
"past_encoder_ctx": past_encoder_ctx,
}
return (
xs_pad.new_zeros(bsize, 0, self._output_size),
xs_pad.new_zeros(bsize),
next_states,
)
overlap_size = self.block_size - self.hop_size
block_num = max(0, xs_pad.size(1) - overlap_size) // self.hop_size
res_frame_num = xs_pad.size(1) - self.hop_size * block_num
buffer_after_downsampling = xs_pad.narrow(
1, xs_pad.size(1) - res_frame_num, res_frame_num
)
xs_pad = xs_pad.narrow(1, 0, block_num * self.hop_size + overlap_size)
# block_size could be 0 meaning infinite
# apply usual encoder for short sequence
assert self.block_size > 0
if n_processed_blocks == 0 and total_frame_num <= self.block_size and is_final:
xs_chunk = self.pos_enc(xs_pad).unsqueeze(1)
xs_pad, _, _, _, _, _, _ = self.encoders(
xs_chunk, None, True, None, None, True
)
xs_pad = xs_pad.squeeze(0)
if self.normalize_before:
xs_pad = self.after_norm(xs_pad)
return xs_pad, None, None
# start block processing
xs_chunk = xs_pad.new_zeros(
bsize, block_num, self.block_size + 2, xs_pad.size(-1)
)
for i in range(block_num):
cur_hop = i * self.hop_size
chunk_length = min(self.block_size, total_frame_num - cur_hop)
addin = xs_pad.narrow(1, cur_hop, chunk_length)
if self.init_average:
addin = addin.mean(1, keepdim=True)
else:
addin = addin.max(1, keepdim=True)
if self.ctx_pos_enc:
addin = self.pos_enc(addin, i + n_processed_blocks)
if prev_addin is None:
prev_addin = addin
xs_chunk[:, i, 0] = prev_addin
xs_chunk[:, i, -1] = addin
chunk = self.pos_enc(
xs_pad.narrow(1, cur_hop, chunk_length),
cur_hop + self.hop_size * n_processed_blocks,
)
xs_chunk[:, i, 1 : chunk_length + 1] = chunk
prev_addin = addin
# mask setup, it should be the same to that of forward_train
mask_online = xs_pad.new_zeros(
xs_pad.size(0), block_num, self.block_size + 2, self.block_size + 2
)
mask_online.narrow(2, 1, self.block_size + 1).narrow(
3, 0, self.block_size + 1
).fill_(1)
ys_chunk, _, _, _, past_encoder_ctx, _, _ = self.encoders(
xs_chunk, mask_online, True, past_encoder_ctx
)
# remove addin
ys_chunk = ys_chunk.narrow(2, 1, self.block_size)
offset = self.block_size - self.look_ahead - self.hop_size
if is_final:
if n_processed_blocks == 0:
y_length = xs_pad.size(1)
else:
y_length = xs_pad.size(1) - offset
else:
y_length = block_num * self.hop_size
if n_processed_blocks == 0:
y_length += offset
ys_pad = xs_pad.new_zeros((xs_pad.size(0), y_length, xs_pad.size(2)))
if n_processed_blocks == 0:
ys_pad[:, 0:offset] = ys_chunk[:, 0, 0:offset]
for i in range(block_num):
cur_hop = i * self.hop_size
if n_processed_blocks == 0:
cur_hop += offset
if i == block_num - 1 and is_final:
chunk_length = min(self.block_size - offset, ys_pad.size(1) - cur_hop)
else:
chunk_length = self.hop_size
ys_pad[:, cur_hop : cur_hop + chunk_length] = ys_chunk[
:, i, offset : offset + chunk_length
]
if self.normalize_before:
ys_pad = self.after_norm(ys_pad)
if is_final:
next_states = None
else:
next_states = {
"prev_addin": prev_addin,
"buffer_before_downsampling": buffer_before_downsampling,
"ilens_buffer": ilens_buffer,
"buffer_after_downsampling": buffer_after_downsampling,
"n_processed_blocks": n_processed_blocks + block_num,
"past_encoder_ctx": past_encoder_ctx,
}
return ys_pad, None, next_states