tuna2134/sbv2-v2
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

fix dur

Browse Source
This commit is contained in:
Stardust·减
2023-09-03 14:23:04 +08:00
committed by GitHub
parent d0015a0ca4
commit ca4b39cdcd
1 changed files with 21 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files

42
models.py
View File

@@ -26,18 +26,18 @@ class DurationDiscriminator(nn.Module): #vits2
self.drop = nn.Dropout(p_dropout) self.drop = nn.Dropout(p_dropout)
self.conv_1 = nn.Conv1d(in_channels, filter_channels, kernel_size, padding=kernel_size//2) self.conv_1 = nn.Conv1d(in_channels, filter_channels, kernel_size, padding=kernel_size//2)
self.norm_1 = modules.LayerNorm(filter_channels) #self.norm_1 = modules.LayerNorm(filter_channels)
self.conv_2 = nn.Conv1d(filter_channels, filter_channels, kernel_size, padding=kernel_size//2) self.conv_2 = nn.Conv1d(filter_channels, filter_channels, kernel_size, padding=kernel_size//2)
self.norm_2 = modules.LayerNorm(filter_channels) #self.norm_2 = modules.LayerNorm(filter_channels)
self.dur_proj = nn.Conv1d(1, filter_channels, 1) self.dur_proj = nn.Conv1d(1, filter_channels, 1)
self.pre_out_conv_1 = nn.Conv1d(2*filter_channels, filter_channels, kernel_size, padding=kernel_size//2) self.pre_out_conv_1 = nn.Conv1d(2*filter_channels, filter_channels, kernel_size, padding=kernel_size//2)
self.pre_out_norm_1 = modules.LayerNorm(filter_channels) #self.pre_out_norm_1 = modules.LayerNorm(filter_channels)
self.pre_out_conv_2 = nn.Conv1d(filter_channels, filter_channels, kernel_size, padding=kernel_size//2) self.pre_out_conv_2 = nn.Conv1d(filter_channels, filter_channels, kernel_size, padding=kernel_size//2)
self.pre_out_norm_2 = modules.LayerNorm(filter_channels) #self.pre_out_norm_2 = modules.LayerNorm(filter_channels)
if gin_channels != 0: #if gin_channels != 0:
self.cond = nn.Conv1d(gin_channels, in_channels, 1) # self.cond = nn.Conv1d(gin_channels, in_channels, 1)
self.output_layer = nn.Sequential( self.output_layer = nn.Sequential(
nn.Linear(filter_channels, 1), nn.Linear(filter_channels, 1),
@@ -48,13 +48,13 @@ class DurationDiscriminator(nn.Module): #vits2
dur = self.dur_proj(dur) dur = self.dur_proj(dur)
x = torch.cat([x, dur], dim=1) x = torch.cat([x, dur], dim=1)
x = self.pre_out_conv_1(x * x_mask) x = self.pre_out_conv_1(x * x_mask)
x = torch.relu(x) #x = torch.relu(x)
x = self.pre_out_norm_1(x) #x = self.pre_out_norm_1(x)
x = self.drop(x) #x = self.drop(x)
x = self.pre_out_conv_2(x * x_mask) x = self.pre_out_conv_2(x * x_mask)
x = torch.relu(x) #x = torch.relu(x)
x = self.pre_out_norm_2(x) #x = self.pre_out_norm_2(x)
x = self.drop(x) #x = self.drop(x)
x = x * x_mask x = x * x_mask
x = x.transpose(1, 2) x = x.transpose(1, 2)
output_prob = self.output_layer(x) output_prob = self.output_layer(x)
@@ -62,17 +62,17 @@ class DurationDiscriminator(nn.Module): #vits2
def forward(self, x, x_mask, dur_r, dur_hat, g=None): def forward(self, x, x_mask, dur_r, dur_hat, g=None):
x = torch.detach(x) x = torch.detach(x)
if g is not None: #if g is not None:
g = torch.detach(g) # g = torch.detach(g)
x = x + self.cond(g) # x = x + self.cond(g)
x = self.conv_1(x * x_mask) x = self.conv_1(x * x_mask)
x = torch.relu(x) #x = torch.relu(x)
x = self.norm_1(x) #x = self.norm_1(x)
x = self.drop(x) #x = self.drop(x)
x = self.conv_2(x * x_mask) x = self.conv_2(x * x_mask)
x = torch.relu(x) #x = torch.relu(x)
x = self.norm_2(x) #x = self.norm_2(x)
x = self.drop(x) #x = self.drop(x)
output_probs = [] output_probs = []
for dur in [dur_r, dur_hat]: for dur in [dur_r, dur_hat]: