449 lines
16 KiB
Python
449 lines
16 KiB
Python
import torch
|
|
import torch.nn.functional as F
|
|
import torch.nn as nn
|
|
from torch.nn import Conv1d, ConvTranspose1d, AvgPool1d, Conv2d
|
|
from torch.nn.utils import weight_norm, remove_weight_norm, spectral_norm
|
|
from models.mossformer2_sr.utils import init_weights, get_padding
|
|
from models.mossformer2_sr.mossformer2 import MossFormer_MaskNet
|
|
from models.mossformer2_sr.snake import Snake1d
|
|
from typing import Optional, List, Union, Dict, Tuple
|
|
from models.mossformer2_sr.env import AttrDict
|
|
import typing
|
|
from torchaudio.transforms import Spectrogram, Resample
|
|
|
|
LRELU_SLOPE = 0.1
|
|
|
|
|
|
class ResBlock1(torch.nn.Module):
|
|
def __init__(self, h, channels, kernel_size=3, dilation=(1, 3, 5)):
|
|
super(ResBlock1, self).__init__()
|
|
self.h = h
|
|
self.convs1 = nn.ModuleList([
|
|
weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=dilation[0],
|
|
padding=get_padding(kernel_size, dilation[0]))),
|
|
weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=dilation[1],
|
|
padding=get_padding(kernel_size, dilation[1]))),
|
|
weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=dilation[2],
|
|
padding=get_padding(kernel_size, dilation[2])))
|
|
#Snake1d(channels)
|
|
])
|
|
self.convs1.apply(init_weights)
|
|
self.convs1_activates = nn.ModuleList([
|
|
Snake1d(channels),
|
|
Snake1d(channels),
|
|
Snake1d(channels)
|
|
])
|
|
self.convs2 = nn.ModuleList([
|
|
weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=1,
|
|
padding=get_padding(kernel_size, 1))),
|
|
weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=1,
|
|
padding=get_padding(kernel_size, 1))),
|
|
weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=1,
|
|
padding=get_padding(kernel_size, 1)))
|
|
#Snake1d(channels)
|
|
])
|
|
self.convs2.apply(init_weights)
|
|
#self.convs2_activate = Snake1d(channels)
|
|
self.convs2_activates = nn.ModuleList([
|
|
Snake1d(channels),
|
|
Snake1d(channels),
|
|
Snake1d(channels)
|
|
])
|
|
|
|
def forward(self, x):
|
|
for c1, c2, act1, act2 in zip(self.convs1, self.convs2, self.convs1_activates, self.convs2_activates):
|
|
#xt = F.leaky_relu(x, LRELU_SLOPE)
|
|
#print(f'xt: {xt.shape}')
|
|
xt = act1(x)
|
|
xt = c1(xt)
|
|
#xt = F.leaky_relu(xt, LRELU_SLOPE)
|
|
xt = act2(xt)
|
|
xt = c2(xt)
|
|
x = xt + x
|
|
return x
|
|
|
|
def remove_weight_norm(self):
|
|
for l in self.convs1:
|
|
remove_weight_norm(l)
|
|
for l in self.convs2:
|
|
remove_weight_norm(l)
|
|
|
|
|
|
class ResBlock2(torch.nn.Module):
|
|
def __init__(self, h, channels, kernel_size=3, dilation=(1, 3)):
|
|
super(ResBlock2, self).__init__()
|
|
self.h = h
|
|
self.convs = nn.ModuleList([
|
|
weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=dilation[0],
|
|
padding=get_padding(kernel_size, dilation[0]))),
|
|
weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=dilation[1],
|
|
padding=get_padding(kernel_size, dilation[1])))
|
|
#Snake1d(channels)
|
|
])
|
|
self.convs.apply(init_weights)
|
|
#self.convs_activate = Snake1d(channels)
|
|
self.convs_activates = nn.ModuleList([
|
|
Snake1d(channels),
|
|
Snake1d(channels)
|
|
])
|
|
def forward(self, x):
|
|
for c, act in zip(self.convs, self.convs_activates):
|
|
#xt = F.leaky_relu(x, LRELU_SLOPE)
|
|
xt = act(x)
|
|
xt = c(xt)
|
|
x = xt + x
|
|
return x
|
|
|
|
def remove_weight_norm(self):
|
|
for l in self.convs:
|
|
remove_weight_norm(l)
|
|
|
|
|
|
class Generator(torch.nn.Module):
|
|
def __init__(self, h):
|
|
super(Generator, self).__init__()
|
|
self.h = h
|
|
self.num_kernels = len(h.resblock_kernel_sizes)
|
|
self.num_upsamples = len(h.upsample_rates)
|
|
self.conv_pre = weight_norm(Conv1d(80, h.upsample_initial_channel, 7, 1, padding=3))
|
|
resblock = ResBlock1 if h.resblock == '1' else ResBlock2
|
|
|
|
self.ups = nn.ModuleList()
|
|
self.snakes = nn.ModuleList()
|
|
for i, (u, k) in enumerate(zip(h.upsample_rates, h.upsample_kernel_sizes)):
|
|
self.snakes.append(Snake1d(h.upsample_initial_channel//(2**i)))
|
|
self.ups.append(weight_norm(
|
|
ConvTranspose1d(h.upsample_initial_channel//(2**i), h.upsample_initial_channel//(2**(i+1)),
|
|
k, u, padding=(k-u)//2)))
|
|
|
|
self.resblocks = nn.ModuleList()
|
|
for i in range(len(self.ups)):
|
|
ch = h.upsample_initial_channel//(2**(i+1))
|
|
for j, (k, d) in enumerate(zip(h.resblock_kernel_sizes, h.resblock_dilation_sizes)):
|
|
self.resblocks.append(resblock(h, ch, k, d))
|
|
|
|
self.snake_post = Snake1d(ch)
|
|
self.conv_post = weight_norm(Conv1d(ch, 1, 7, 1, padding=3))
|
|
self.ups.apply(init_weights)
|
|
self.conv_post.apply(init_weights)
|
|
def forward(self, x):
|
|
x = self.conv_pre(x)
|
|
for i in range(self.num_upsamples):
|
|
#x = F.leaky_relu(x, LRELU_SLOPE)
|
|
#print(f'x {i}: {x.shape}')
|
|
x = self.snakes[i](x)
|
|
x = self.ups[i](x)
|
|
xs = None
|
|
for j in range(self.num_kernels):
|
|
if xs is None:
|
|
xs = self.resblocks[i*self.num_kernels+j](x)
|
|
else:
|
|
xs += self.resblocks[i*self.num_kernels+j](x)
|
|
x = xs / self.num_kernels
|
|
#x = F.leaky_relu(x)
|
|
x = self.snake_post(x)
|
|
x = self.conv_post(x)
|
|
x = torch.tanh(x)
|
|
|
|
return x
|
|
|
|
def remove_weight_norm(self):
|
|
#print('Removing weight norm...')
|
|
for l in self.ups:
|
|
remove_weight_norm(l)
|
|
for l in self.resblocks:
|
|
l.remove_weight_norm()
|
|
remove_weight_norm(self.conv_pre)
|
|
remove_weight_norm(self.conv_post)
|
|
|
|
|
|
class DiscriminatorP(torch.nn.Module):
|
|
def __init__(self, period, kernel_size=5, stride=3, use_spectral_norm=False):
|
|
super(DiscriminatorP, self).__init__()
|
|
self.period = period
|
|
norm_f = weight_norm if use_spectral_norm == False else spectral_norm
|
|
self.convs = nn.ModuleList([
|
|
norm_f(Conv2d(1, 32, (kernel_size, 1), (stride, 1), padding=(get_padding(5, 1), 0))),
|
|
norm_f(Conv2d(32, 128, (kernel_size, 1), (stride, 1), padding=(get_padding(5, 1), 0))),
|
|
norm_f(Conv2d(128, 512, (kernel_size, 1), (stride, 1), padding=(get_padding(5, 1), 0))),
|
|
norm_f(Conv2d(512, 1024, (kernel_size, 1), (stride, 1), padding=(get_padding(5, 1), 0))),
|
|
norm_f(Conv2d(1024, 1024, (kernel_size, 1), 1, padding=(2, 0))),
|
|
])
|
|
self.conv_post = norm_f(Conv2d(1024, 1, (3, 1), 1, padding=(1, 0)))
|
|
|
|
def forward(self, x):
|
|
fmap = []
|
|
|
|
# 1d to 2d
|
|
b, c, t = x.shape
|
|
if t % self.period != 0: # pad first
|
|
n_pad = self.period - (t % self.period)
|
|
x = F.pad(x, (0, n_pad), "reflect")
|
|
t = t + n_pad
|
|
x = x.view(b, c, t // self.period, self.period)
|
|
|
|
for l in self.convs:
|
|
x = l(x)
|
|
x = F.leaky_relu(x, LRELU_SLOPE)
|
|
fmap.append(x)
|
|
x = self.conv_post(x)
|
|
fmap.append(x)
|
|
x = torch.flatten(x, 1, -1)
|
|
|
|
return x, fmap
|
|
|
|
|
|
class MultiPeriodDiscriminator(torch.nn.Module):
|
|
def __init__(self):
|
|
super(MultiPeriodDiscriminator, self).__init__()
|
|
self.discriminators = nn.ModuleList([
|
|
DiscriminatorP(2),
|
|
DiscriminatorP(3),
|
|
DiscriminatorP(5),
|
|
DiscriminatorP(7),
|
|
DiscriminatorP(11),
|
|
])
|
|
|
|
def forward(self, y, y_hat):
|
|
y_d_rs = []
|
|
y_d_gs = []
|
|
fmap_rs = []
|
|
fmap_gs = []
|
|
for i, d in enumerate(self.discriminators):
|
|
y_d_r, fmap_r = d(y)
|
|
y_d_g, fmap_g = d(y_hat)
|
|
y_d_rs.append(y_d_r)
|
|
fmap_rs.append(fmap_r)
|
|
y_d_gs.append(y_d_g)
|
|
fmap_gs.append(fmap_g)
|
|
|
|
return y_d_rs, y_d_gs, fmap_rs, fmap_gs
|
|
|
|
|
|
class DiscriminatorS(torch.nn.Module):
|
|
def __init__(self, use_spectral_norm=False):
|
|
super(DiscriminatorS, self).__init__()
|
|
norm_f = weight_norm if use_spectral_norm == False else spectral_norm
|
|
self.convs = nn.ModuleList([
|
|
norm_f(Conv1d(1, 128, 15, 1, padding=7)),
|
|
norm_f(Conv1d(128, 128, 41, 2, groups=4, padding=20)),
|
|
norm_f(Conv1d(128, 256, 41, 2, groups=16, padding=20)),
|
|
norm_f(Conv1d(256, 512, 41, 4, groups=16, padding=20)),
|
|
norm_f(Conv1d(512, 1024, 41, 4, groups=16, padding=20)),
|
|
norm_f(Conv1d(1024, 1024, 41, 1, groups=16, padding=20)),
|
|
norm_f(Conv1d(1024, 1024, 5, 1, padding=2)),
|
|
])
|
|
self.conv_post = norm_f(Conv1d(1024, 1, 3, 1, padding=1))
|
|
|
|
def forward(self, x):
|
|
fmap = []
|
|
for l in self.convs:
|
|
x = l(x)
|
|
x = F.leaky_relu(x, LRELU_SLOPE)
|
|
fmap.append(x)
|
|
x = self.conv_post(x)
|
|
fmap.append(x)
|
|
x = torch.flatten(x, 1, -1)
|
|
|
|
return x, fmap
|
|
|
|
|
|
class MultiScaleDiscriminator(torch.nn.Module):
|
|
def __init__(self):
|
|
super(MultiScaleDiscriminator, self).__init__()
|
|
self.discriminators = nn.ModuleList([
|
|
DiscriminatorS(use_spectral_norm=True),
|
|
DiscriminatorS(),
|
|
DiscriminatorS(),
|
|
])
|
|
self.meanpools = nn.ModuleList([
|
|
AvgPool1d(4, 2, padding=2),
|
|
AvgPool1d(4, 2, padding=2)
|
|
])
|
|
|
|
def forward(self, y, y_hat):
|
|
y_d_rs = []
|
|
y_d_gs = []
|
|
fmap_rs = []
|
|
fmap_gs = []
|
|
for i, d in enumerate(self.discriminators):
|
|
if i != 0:
|
|
y = self.meanpools[i-1](y)
|
|
y_hat = self.meanpools[i-1](y_hat)
|
|
y_d_r, fmap_r = d(y)
|
|
y_d_g, fmap_g = d(y_hat)
|
|
y_d_rs.append(y_d_r)
|
|
fmap_rs.append(fmap_r)
|
|
y_d_gs.append(y_d_g)
|
|
fmap_gs.append(fmap_g)
|
|
|
|
return y_d_rs, y_d_gs, fmap_rs, fmap_gs
|
|
|
|
|
|
# Method based on descript-audio-codec: https://github.com/descriptinc/descript-audio-codec
|
|
# Modified code adapted from https://github.com/gemelo-ai/vocos under the MIT license.
|
|
# LICENSE is in incl_licenses directory.
|
|
class DiscriminatorB(nn.Module):
|
|
def __init__(
|
|
self,
|
|
window_length: int,
|
|
channels: int = 32,
|
|
hop_factor: float = 0.25,
|
|
bands: Tuple[Tuple[float, float], ...] = (
|
|
(0.0, 0.1),
|
|
(0.1, 0.25),
|
|
(0.25, 0.5),
|
|
(0.5, 0.75),
|
|
(0.75, 1.0),
|
|
),
|
|
):
|
|
super().__init__()
|
|
self.window_length = window_length
|
|
self.hop_factor = hop_factor
|
|
self.spec_fn = Spectrogram(
|
|
n_fft=window_length,
|
|
hop_length=int(window_length * hop_factor),
|
|
win_length=window_length,
|
|
power=None,
|
|
)
|
|
n_fft = window_length // 2 + 1
|
|
bands = [(int(b[0] * n_fft), int(b[1] * n_fft)) for b in bands]
|
|
self.bands = bands
|
|
convs = lambda: nn.ModuleList(
|
|
[
|
|
weight_norm(nn.Conv2d(2, channels, (3, 9), (1, 1), padding=(1, 4))),
|
|
weight_norm(
|
|
nn.Conv2d(channels, channels, (3, 9), (1, 2), padding=(1, 4))
|
|
),
|
|
weight_norm(
|
|
nn.Conv2d(channels, channels, (3, 9), (1, 2), padding=(1, 4))
|
|
),
|
|
weight_norm(
|
|
nn.Conv2d(channels, channels, (3, 9), (1, 2), padding=(1, 4))
|
|
),
|
|
weight_norm(
|
|
nn.Conv2d(channels, channels, (3, 3), (1, 1), padding=(1, 1))
|
|
),
|
|
]
|
|
)
|
|
self.band_convs = nn.ModuleList([convs() for _ in range(len(self.bands))])
|
|
|
|
self.conv_post = weight_norm(
|
|
nn.Conv2d(channels, 1, (3, 3), (1, 1), padding=(1, 1))
|
|
)
|
|
def spectrogram(self, x: torch.Tensor) -> List[torch.Tensor]:
|
|
# Remove DC offset
|
|
x = x - x.mean(dim=-1, keepdims=True)
|
|
# Peak normalize the volume of input audio
|
|
x = 0.8 * x / (x.abs().max(dim=-1, keepdim=True)[0] + 1e-9)
|
|
x = self.spec_fn(x)
|
|
x = torch.view_as_real(x)
|
|
x = x.permute(0, 3, 2, 1) # [B, F, T, C] -> [B, C, T, F]
|
|
# Split into bands
|
|
x_bands = [x[..., b[0] : b[1]] for b in self.bands]
|
|
return x_bands
|
|
|
|
def forward(self, x: torch.Tensor) -> Tuple[torch.Tensor, List[torch.Tensor]]:
|
|
x_bands = self.spectrogram(x.squeeze(1))
|
|
fmap = []
|
|
x = []
|
|
|
|
for band, stack in zip(x_bands, self.band_convs):
|
|
for i, layer in enumerate(stack):
|
|
band = layer(band)
|
|
band = torch.nn.functional.leaky_relu(band, 0.1)
|
|
if i > 0:
|
|
fmap.append(band)
|
|
x.append(band)
|
|
|
|
x = torch.cat(x, dim=-1)
|
|
x = self.conv_post(x)
|
|
fmap.append(x)
|
|
|
|
return x, fmap
|
|
|
|
# Method based on descript-audio-codec: https://github.com/descriptinc/descript-audio-codec
|
|
# Modified code adapted from https://github.com/gemelo-ai/vocos under the MIT license.
|
|
# LICENSE is in incl_licenses directory.
|
|
class MultiBandDiscriminator(nn.Module):
|
|
def __init__(
|
|
self,
|
|
h,
|
|
):
|
|
"""
|
|
Multi-band multi-scale STFT discriminator, with the architecture based on https://github.com/descriptinc/descript-audio-codec.
|
|
and the modified code adapted from https://github.com/gemelo-ai/vocos.
|
|
"""
|
|
super().__init__()
|
|
# fft_sizes (list[int]): Tuple of window lengths for FFT. Defaults to [2048, 1024, 512] if not set in h.
|
|
self.fft_sizes = h.get("mbd_fft_sizes", [2048, 1024, 512])
|
|
self.discriminators = nn.ModuleList(
|
|
[DiscriminatorB(window_length=w) for w in self.fft_sizes]
|
|
)
|
|
|
|
def forward(self, y: torch.Tensor, y_hat: torch.Tensor) -> Tuple[
|
|
List[torch.Tensor],
|
|
List[torch.Tensor],
|
|
List[List[torch.Tensor]],
|
|
List[List[torch.Tensor]],
|
|
]:
|
|
|
|
y_d_rs = []
|
|
y_d_gs = []
|
|
fmap_rs = []
|
|
fmap_gs = []
|
|
|
|
for d in self.discriminators:
|
|
y_d_r, fmap_r = d(x=y)
|
|
y_d_g, fmap_g = d(x=y_hat)
|
|
y_d_rs.append(y_d_r)
|
|
fmap_rs.append(fmap_r)
|
|
y_d_gs.append(y_d_g)
|
|
fmap_gs.append(fmap_g)
|
|
|
|
return y_d_rs, y_d_gs, fmap_rs, fmap_gs
|
|
|
|
def feature_loss(fmap_r, fmap_g):
|
|
loss = 0
|
|
for dr, dg in zip(fmap_r, fmap_g):
|
|
for rl, gl in zip(dr, dg):
|
|
loss += torch.mean(torch.abs(rl - gl))
|
|
|
|
return loss*2
|
|
|
|
|
|
def discriminator_loss(disc_real_outputs, disc_generated_outputs):
|
|
loss = 0
|
|
r_losses = []
|
|
g_losses = []
|
|
for dr, dg in zip(disc_real_outputs, disc_generated_outputs):
|
|
r_loss = torch.mean((1-dr)**2)
|
|
g_loss = torch.mean(dg**2)
|
|
loss += (r_loss + g_loss)
|
|
r_losses.append(r_loss.item())
|
|
g_losses.append(g_loss.item())
|
|
|
|
return loss, r_losses, g_losses
|
|
|
|
|
|
def generator_loss(disc_outputs):
|
|
loss = 0
|
|
gen_losses = []
|
|
for dg in disc_outputs:
|
|
l = torch.mean((1-dg)**2)
|
|
gen_losses.append(l)
|
|
loss += l
|
|
|
|
return loss, gen_losses
|
|
|
|
class Mossformer(nn.Module):
|
|
|
|
def __init__(self):
|
|
super(Mossformer, self).__init__()
|
|
self.mossformer = MossFormer_MaskNet(in_channels=80, out_channels=512, out_channels_final=80)
|
|
|
|
def forward(self, input):
|
|
out = self.mossformer(input)
|
|
return out
|
|
|