src/kanade_tokenizer/module/convnext.py

# Adapted from: https://github.com/gemelo-ai/vocos/blob/main/vocos/models.py


import torch
from torch import nn


class ConvNeXtBlock(nn.Module):
    """ConvNeXt Block adapted from https://github.com/facebookresearch/ConvNeXt to 1D audio signal.

    Args:
        dim (int): Number of input channels.
        intermediate_dim (int): Dimensionality of the intermediate layer.
        layer_scale_init_value (float, optional): Initial value for the layer scale. None means no scaling.
            Defaults to None.
    """

    def __init__(
        self,
        dim: int,
        intermediate_dim: int,
        layer_scale_init_value: float,
    ):
        super().__init__()
        self.dwconv = nn.Conv1d(dim, dim, kernel_size=7, padding=3, groups=dim)  # depthwise conv
        self.norm = nn.LayerNorm(dim, eps=1e-6)
        self.pwconv1 = nn.Linear(dim, intermediate_dim)  # pointwise/1x1 convs, implemented with linear layers
        self.act = nn.GELU()
        self.pwconv2 = nn.Linear(intermediate_dim, dim)
        self.gamma = (
            nn.Parameter(layer_scale_init_value * torch.ones(dim), requires_grad=True)
            if layer_scale_init_value > 0
            else None
        )

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        residual = x
        x = self.dwconv(x)
        x = x.transpose(1, 2)  # (B, C, T) -> (B, T, C)
        x = self.norm(x)
        x = self.pwconv1(x)
        x = self.act(x)
        x = self.pwconv2(x)
        if self.gamma is not None:
            x = self.gamma * x
        x = x.transpose(1, 2)  # (B, T, C) -> (B, C, T)

        x = residual + x
        return x


class ConvNextBackbone(nn.Module):
    """
    Backbone module built with ConvNeXt blocks.

    Args:
        input_channels (int): Number of input features channels.
        dim (int): Hidden dimension of the model.
        intermediate_dim (int): Intermediate dimension used in ConvNeXtBlock.
        num_layers (int): Number of ConvNeXtBlock layers.
        layer_scale_init_value (float, optional): Initial value for layer scaling. Defaults to `1 / num_layers`.
    """

    def __init__(
        self,
        input_channels: int,
        dim: int,
        intermediate_dim: int,
        num_layers: int,
        output_channels: int | None = None,
        layer_scale_init_value: float | None = None,
        skip_embed: bool = False,
    ):
        super().__init__()
        self.input_channels = input_channels
        self.output_channels = output_channels
        self.dim = dim
        self.embed = nn.Conv1d(input_channels, dim, kernel_size=7, padding=3) if not skip_embed else nn.Identity()
        self.norm = nn.LayerNorm(dim, eps=1e-6)
        layer_scale_init_value = layer_scale_init_value or 1 / num_layers
        self.convnext = nn.ModuleList(
            [
                ConvNeXtBlock(
                    dim=dim,
                    intermediate_dim=intermediate_dim,
                    layer_scale_init_value=layer_scale_init_value,
                )
                for _ in range(num_layers)
            ]
        )
        self.proj_out = nn.Linear(dim, output_channels) if output_channels else nn.Identity()
        self.final_layer_norm = nn.LayerNorm(dim, eps=1e-6)
        self.apply(self._init_weights)

    @property
    def input_dim(self) -> int:
        return self.input_channels

    @property
    def output_dim(self) -> int:
        return self.output_channels if self.output_channels else self.dim

    def _init_weights(self, m):
        if isinstance(m, (nn.Conv1d, nn.Linear)):
            nn.init.trunc_normal_(m.weight, std=0.02)
            nn.init.constant_(m.bias, 0)

    def forward(self, x: torch.Tensor, **kwargs) -> torch.Tensor:
        """
        Args:
            x (Tensor): Input tensor of shape (B, L, C), where B is the batch size,
                        C denotes output features, and L is the sequence length.
        Returns:
            Tensor: Output of shape (B, L, H), where B is the batch size, L is the sequence length,
                    and H denotes the model dimension.
        """
        x = x.transpose(1, 2)  # (B, L, C) -> (B, C, L)
        x = self.embed(x)
        x = self.norm(x.transpose(1, 2))
        x = x.transpose(1, 2)
        for conv_block in self.convnext:
            x = conv_block(x)
        x = self.final_layer_norm(x.transpose(1, 2))
        x = self.proj_out(x)
        return x