Source code for openspeech.modules.conformer_block

# MIT License
#
# Copyright (c) 2021 Soohwan Kim and Sangchun Ha and Soyoung Cho
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in all
# copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.

import torch.nn as nn
from torch import Tensor

from openspeech.modules.conformer_attention_module import MultiHeadedSelfAttentionModule
from openspeech.modules.conformer_convolution_module import ConformerConvModule
from openspeech.modules.conformer_feed_forward_module import FeedForwardModule
from openspeech.modules.residual_connection_module import ResidualConnectionModule


class ConformerBlock(nn.Module):
    r"""
    Conformer block contains two Feed Forward modules sandwiching the Multi-Headed Self-Attention module
    and the Convolution module. This sandwich structure is inspired by Macaron-Net, which proposes replacing
    the original feed-forward layer in the Transformer block into two half-step feed-forward layers,
    one before the attention layer and one after.

    Args:
        encoder_dim (int, optional): Dimension of conformer encoders
        num_attention_heads (int, optional): Number of attention heads
        feed_forward_expansion_factor (int, optional): Expansion factor of feed forward module
        conv_expansion_factor (int, optional): Expansion factor of conformer convolution module
        feed_forward_dropout_p (float, optional): Probability of feed forward module dropout
        attention_dropout_p (float, optional): Probability of attention module dropout
        conv_dropout_p (float, optional): Probability of conformer convolution module dropout
        conv_kernel_size (int or tuple, optional): Size of the convolving kernel
        half_step_residual (bool): Flag indication whether to use half step residual or not

    Inputs: inputs
        - **inputs** (batch, time, dim): Tensor containing input vector

    Returns: outputs
        - **outputs** (batch, time, dim): Tensor produces by conformer block.
    """
    def __init__(
            self,
            encoder_dim: int = 512,
            num_attention_heads: int = 8,
            feed_forward_expansion_factor: int = 4,
            conv_expansion_factor: int = 2,
            feed_forward_dropout_p: float = 0.1,
            attention_dropout_p: float = 0.1,
            conv_dropout_p: float = 0.1,
            conv_kernel_size: int = 31,
            half_step_residual: bool = True,
    ) -> None:
        super(ConformerBlock, self).__init__()
        if half_step_residual:
            self.feed_forward_residual_factor = 0.5
        else:
            self.feed_forward_residual_factor = 1

        self.sequential = nn.Sequential(
            ResidualConnectionModule(
                module=FeedForwardModule(
                    encoder_dim=encoder_dim,
                    expansion_factor=feed_forward_expansion_factor,
                    dropout_p=feed_forward_dropout_p,
                ),
                module_factor=self.feed_forward_residual_factor,
            ),
            ResidualConnectionModule(
                module=MultiHeadedSelfAttentionModule(
                    d_model=encoder_dim,
                    num_heads=num_attention_heads,
                    dropout_p=attention_dropout_p,
                ),
            ),
            ResidualConnectionModule(
                module=ConformerConvModule(
                    in_channels=encoder_dim,
                    kernel_size=conv_kernel_size,
                    expansion_factor=conv_expansion_factor,
                    dropout_p=conv_dropout_p,
                ),
            ),
            ResidualConnectionModule(
                module=FeedForwardModule(
                    encoder_dim=encoder_dim,
                    expansion_factor=feed_forward_expansion_factor,
                    dropout_p=feed_forward_dropout_p,
                ),
                module_factor=self.feed_forward_residual_factor,
            ),
            nn.LayerNorm(encoder_dim),
        )

    def forward(self, inputs: Tensor) -> Tensor:
        return self.sequential(inputs)