DeepLearningExamples/PyTorch/SpeechSynthesis/HiFiGAN/common/layers.py

# *****************************************************************************
#  Copyright (c) 2018, NVIDIA CORPORATION.  All rights reserved.
#
#  Redistribution and use in source and binary forms, with or without
#  modification, are permitted provided that the following conditions are met:
#      * Redistributions of source code must retain the above copyright
#        notice, this list of conditions and the following disclaimer.
#      * Redistributions in binary form must reproduce the above copyright
#        notice, this list of conditions and the following disclaimer in the
#        documentation and/or other materials provided with the distribution.
#      * Neither the name of the NVIDIA CORPORATION nor the
#        names of its contributors may be used to endorse or promote products
#        derived from this software without specific prior written permission.
#
#  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
#  ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
#  WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
#  DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
#  DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
#  (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
#  LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
#  ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
#  (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
#  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#
# *****************************************************************************

import torch
import torch.nn.functional as F
from librosa.filters import mel as librosa_mel_fn

from common.audio_processing import (dynamic_range_compression,
                                     dynamic_range_decompression)
from common.stft import STFT


class LinearNorm(torch.nn.Module):
    def __init__(self, in_dim, out_dim, bias=True, w_init_gain='linear'):
        super(LinearNorm, self).__init__()
        self.linear_layer = torch.nn.Linear(in_dim, out_dim, bias=bias)

        torch.nn.init.xavier_uniform_(
            self.linear_layer.weight,
            gain=torch.nn.init.calculate_gain(w_init_gain))

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


class ConvNorm(torch.nn.Module):
    def __init__(self, in_channels, out_channels, kernel_size=1, stride=1,
                 padding=None, dilation=1, bias=True, w_init_gain='linear',
                 batch_norm=False):
        super(ConvNorm, self).__init__()
        if padding is None:
            assert(kernel_size % 2 == 1)
            padding = int(dilation * (kernel_size - 1) / 2)

        self.conv = torch.nn.Conv1d(in_channels, out_channels,
                                    kernel_size=kernel_size, stride=stride,
                                    padding=padding, dilation=dilation,
                                    bias=bias)
        self.norm = torch.nn.BatchNorm1D(out_channels) if batch_norm else None

        torch.nn.init.xavier_uniform_(
            self.conv.weight,
            gain=torch.nn.init.calculate_gain(w_init_gain))

    def forward(self, signal):
        if self.norm is None:
            return self.conv(signal)
        else:
            return self.norm(self.conv(signal))


class ConvReLUNorm(torch.nn.Module):
    def __init__(self, in_channels, out_channels, kernel_size=1, dropout=0.0):
        super(ConvReLUNorm, self).__init__()
        self.conv = torch.nn.Conv1d(in_channels, out_channels,
                                    kernel_size=kernel_size,
                                    padding=(kernel_size // 2))
        self.norm = torch.nn.LayerNorm(out_channels)
        self.dropout = torch.nn.Dropout(dropout)

    def forward(self, signal):
        out = F.relu(self.conv(signal))
        out = self.norm(out.transpose(1, 2)).transpose(1, 2).to(signal.dtype)
        return self.dropout(out)


class TacotronSTFT(torch.nn.Module):
    def __init__(self, filter_length=1024, hop_length=256, win_length=1024,
                 n_mel_channels=80, sampling_rate=22050, mel_fmin=0.0,
                 mel_fmax=8000.0):
        super(TacotronSTFT, self).__init__()
        self.n_mel_channels = n_mel_channels
        self.sampling_rate = sampling_rate
        self.stft_fn = STFT(filter_length, hop_length, win_length)
        mel_basis = librosa_mel_fn(
            sr=sampling_rate,
            n_fft=filter_length,
            n_mels=n_mel_channels,
            fmin=mel_fmin,
            fmax=mel_fmax
        )
        mel_basis = torch.from_numpy(mel_basis).float()
        self.register_buffer('mel_basis', mel_basis)

    def spectral_normalize(self, magnitudes):
        output = dynamic_range_compression(magnitudes)
        return output

    def spectral_de_normalize(self, magnitudes):
        output = dynamic_range_decompression(magnitudes)
        return output

    def mel_spectrogram(self, y):
        """Computes mel-spectrograms from a batch of waves
        PARAMS
        ------
        y: Variable(torch.FloatTensor) with shape (B, T) in range [-1, 1]

        RETURNS
        -------
        mel_output: torch.FloatTensor of shape (B, n_mel_channels, T)
        """
        assert(torch.min(y.data) >= -1)
        assert(torch.max(y.data) <= 1)

        magnitudes, phases = self.stft_fn.transform(y)
        magnitudes = magnitudes.data
        mel_output = torch.matmul(self.mel_basis, magnitudes)
        mel_output = self.spectral_normalize(mel_output)
        return mel_output