matcha_components.py

# Modified from Matcha-TTS https://github.com/shivammehta25/Matcha-TTS
"""
MIT License

Copyright (c) 2023 Shivam Mehta

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 math
from typing import Optional

import torch
import torch.nn as nn
import torch.nn.functional as F

from diffusers.models.activations import get_activation


class SinusoidalPosEmb(torch.nn.Module):
    def __init__(self, dim):
        super().__init__()
        self.dim = dim
        assert self.dim % 2 == 0, "SinusoidalPosEmb requires dim to be even"

    def forward(self, x, scale=1000):
        if x.ndim < 1:
            x = x.unsqueeze(0)
        device = x.device
        half_dim = self.dim // 2
        emb = math.log(10000) / (half_dim - 1)
        emb = torch.exp(torch.arange(half_dim, device=device).float() * -emb)
        emb = scale * x.unsqueeze(1) * emb.unsqueeze(0)
        emb = torch.cat((emb.sin(), emb.cos()), dim=-1)
        return emb


class Block1D(torch.nn.Module):
    def __init__(self, dim, dim_out, groups=8):
        super().__init__()
        self.block = torch.nn.Sequential(
            torch.nn.Conv1d(dim, dim_out, 3, padding=1),
            torch.nn.GroupNorm(groups, dim_out),
            nn.Mish(),
        )

    def forward(self, x, mask):
        output = self.block(x * mask)
        return output * mask


class ResnetBlock1D(torch.nn.Module):
    def __init__(self, dim, dim_out, time_emb_dim, groups=8):
        super().__init__()
        self.mlp = torch.nn.Sequential(
            nn.Mish(), torch.nn.Linear(time_emb_dim, dim_out)
        )

        self.block1 = Block1D(dim, dim_out, groups=groups)
        self.block2 = Block1D(dim_out, dim_out, groups=groups)

        self.res_conv = torch.nn.Conv1d(dim, dim_out, 1)

    def forward(self, x, mask, time_emb):
        h = self.block1(x, mask)
        h += self.mlp(time_emb).unsqueeze(-1)
        h = self.block2(h, mask)
        output = h + self.res_conv(x * mask)
        return output


class Downsample1D(nn.Module):
    def __init__(self, dim):
        super().__init__()
        self.conv = torch.nn.Conv1d(dim, dim, 3, 2, 1)

    def forward(self, x):
        return self.conv(x)


class TimestepEmbedding(nn.Module):
    def __init__(
        self,
        in_channels: int,
        time_embed_dim: int,
        act_fn: str = "silu",
        out_dim: int = None,
        post_act_fn: Optional[str] = None,
        cond_proj_dim=None,
    ):
        super().__init__()

        self.linear_1 = nn.Linear(in_channels, time_embed_dim)

        if cond_proj_dim is not None:
            self.cond_proj = nn.Linear(cond_proj_dim, in_channels, bias=False)
        else:
            self.cond_proj = None

        self.act = get_activation(act_fn)

        if out_dim is not None:
            time_embed_dim_out = out_dim
        else:
            time_embed_dim_out = time_embed_dim
        self.linear_2 = nn.Linear(time_embed_dim, time_embed_dim_out)

        if post_act_fn is None:
            self.post_act = None
        else:
            self.post_act = get_activation(post_act_fn)

    def forward(self, sample, condition=None):
        if condition is not None:
            sample = sample + self.cond_proj(condition)
        sample = self.linear_1(sample)

        if self.act is not None:
            sample = self.act(sample)

        sample = self.linear_2(sample)

        if self.post_act is not None:
            sample = self.post_act(sample)
        return sample


class Upsample1D(nn.Module):
    """A 1D upsampling layer with an optional convolution.

    Parameters:
        channels (`int`):
            number of channels in the inputs and outputs.
        use_conv (`bool`, default `False`):
            option to use a convolution.
        use_conv_transpose (`bool`, default `False`):
            option to use a convolution transpose.
        out_channels (`int`, optional):
            number of output channels. Defaults to `channels`.
    """

    def __init__(
        self,
        channels,
        use_conv=False,
        use_conv_transpose=True,
        out_channels=None,
        name="conv",
    ):
        super().__init__()
        self.channels = channels
        self.out_channels = out_channels or channels
        self.use_conv = use_conv
        self.use_conv_transpose = use_conv_transpose
        self.name = name

        self.conv = None
        if use_conv_transpose:
            self.conv = nn.ConvTranspose1d(channels, self.out_channels, 4, 2, 1)
        elif use_conv:
            self.conv = nn.Conv1d(self.channels, self.out_channels, 3, padding=1)

    def forward(self, inputs):
        assert inputs.shape[1] == self.channels
        if self.use_conv_transpose:
            return self.conv(inputs)

        outputs = F.interpolate(inputs, scale_factor=2.0, mode="nearest")

        if self.use_conv:
            outputs = self.conv(outputs)

        return outputs