def __init__(self):
super(AudioDecoder, self).__init__()
self.hc_blocks = nn.ModuleList([
norm(
ll.CausalConv1d(args.Cx * 2,
args.Cy,
1,
activation_fn=torch.relu))
])
self.hc_blocks.extend([
norm(ll.CausalHighwayConv1d(args.Cy, args.Cy, 3, dilation=3**i))
for i in range(4)
])
self.hc_blocks.extend([
norm(ll.CausalHighwayConv1d(args.Cy, args.Cy, 3, dilation=1))
for _ in range(2)
])
self.hc_blocks.extend([
norm(
ll.CausalConv1d(args.Cy,
args.Cy,
1,
dilation=1,
activation_fn=torch.relu)) for _ in range(3)
])
self.hc_blocks.extend([
norm(ll.CausalConv1d(args.Cy, args.n_mels, 1, dilation=1))
]) # down #filters to dotproduct K, V
def __init__(self,
inChannel,
outChannel,
kernelSize,
padding,
dilation,
activationF=None,
weightNorm=False):
#nn.conv1d(in_channels, out_channels, kernel size, stride, padding,
# dilation, ...)
#
super(Cv, self).__init__()
padDic = {
"same": (kernelSize - 1) * dilation // 2,
"causal": (kernelSize - 1) * dilation,
"none": 0
}
self.pad = padding.lower()
self.padValue = padDic[self.pad]
self.kernelSize = kernelSize
self.dilation = dilation
self.convOne = nn.Conv1d(in_channels=inChannel,
out_channels=outChannel,
kernel_size=kernelSize,
stride=1,
padding=self.padValue,
dilation=dilation)
if weightNorm:
self.convOne = norm(self.convOne)
self.activationF = activationF
self.clear_buffer()
def __init__(self,
inChannel,
outChannel,
kernelSize,
padding,
dilation,
activationF=None,
weightNorm=False):
super(Dc, self).__init__()
padDic = {
"same": dilation * (kernelSize - 1) // 2,
"causal": dilation * (kernelSize - 1),
"none": 0
}
self.pad = padding.lower()
self.padValue = padDic[self.pad]
self.transposedCv = nn.ConvTranspose1d(in_channels=inChannel,
out_channels=outChannel,
kernel_size=kernelSize,
stride=2,
padding=self.padValue,
dilation=dilation)
if weightNorm:
self.convOne = norm(self.convOne)
self.activationF = activationF
def __init__(self):
super(TextEncoder, self).__init__()
self.hc_blocks = nn.ModuleList([
norm(
ll.Conv1d(args.Ce,
args.Cx * 2,
1,
padding='same',
activation_fn=torch.relu))
]) # filter up to split into K, V
self.hc_blocks.extend([
norm(
ll.Conv1d(args.Cx * 2,
args.Cx * 2,
1,
padding='same',
activation_fn=None))
])
self.hc_blocks.extend([
norm(
ll.HighwayConv1d(args.Cx * 2,
args.Cx * 2,
3,
dilation=3**i,
padding='same')) for _ in range(2)
for i in range(4)
])
self.hc_blocks.extend([
norm(
ll.HighwayConv1d(args.Cx * 2,
args.Cx * 2,
3,
dilation=1,
padding='same')) for i in range(2)
])
self.hc_blocks.extend([
norm(
ll.HighwayConv1d(args.Cx * 2,
args.Cx * 2,
1,
dilation=1,
padding='same')) for i in range(2)
])
def __init__(self, d_in, d_out, d_hidden):
super(AudioEncoder, self).__init__()
self.hc_blocks = nn.ModuleList([
norm(ll.CausalConv1d(d_in, d_hidden, 1, activation_fn=torch.relu))
])
self.hc_blocks.extend([
norm(
ll.CausalConv1d(d_hidden,
d_hidden,
1,
activation_fn=torch.relu)) for _ in range(2)
])
self.hc_blocks.extend([
norm(ll.CausalHighwayConv1d(
d_hidden, d_hidden, 3,
dilation=3**i)) # i is in [[0,1,2,3],[0,1,2,3]]
for _ in range(2) for i in range(4)
])
self.hc_blocks.extend([
norm(ll.CausalHighwayConv1d(d_hidden, d_out, 3, dilation=3))
for i in range(2)
])
def __init__(self):
super(AudioEncoder, self).__init__()
self.hc_blocks = nn.ModuleList([
norm(
mm.CausalConv1d(args.n_mels,
args.Cx,
1,
activation_fn=torch.relu))
])
self.hc_blocks.extend([
norm(mm.CausalConv1d(args.Cx, args.Cx, 1,
activation_fn=torch.relu)) for _ in range(2)
])
self.hc_blocks.extend([
norm(mm.CausalHighwayConv1d(
args.Cx, args.Cx, 3,
dilation=3**i)) # i is in [[0,1,2,3],[0,1,2,3]]
for _ in range(2) for i in range(4)
])
self.hc_blocks.extend([
norm(mm.CausalHighwayConv1d(args.Cx, args.Cx, 3, dilation=3))
for i in range(2)
])
def __init__(self):
super(SSRN, self).__init__()
self.name = 'SSRN'
# (N, n_mels, Ty/r) -> (N, Cs, Ty/r)
self.hc_blocks = nn.ModuleList([
norm(ll.Conv1d(args.n_mels, args.Cs, 1, activation_fn=torch.relu))
])
self.hc_blocks.extend([
norm(ll.HighwayConv1d(args.Cs, args.Cs, 3, dilation=3**i))
for i in range(2)
])
# (N, Cs, Ty/r*2) -> (N, Cs, Ty/r*2)
self.hc_blocks.extend([
norm(ll.ConvTranspose1d(args.Cs, args.Cs, 4, stride=2, padding=1))
])
self.hc_blocks.extend([
norm(ll.HighwayConv1d(args.Cs, args.Cs, 3, dilation=3**i))
for i in range(2)
])
# (N, Cs, Ty/r*2) -> (N, Cs, Ty/r*4==Ty)
self.hc_blocks.extend([
norm(ll.ConvTranspose1d(args.Cs, args.Cs, 4, stride=2, padding=1))
])
self.hc_blocks.extend([
norm(ll.HighwayConv1d(args.Cs, args.Cs, 3, dilation=3**i))
for i in range(2)
])
# (N, Cs, Ty) -> (N, Cs*2, Ty)
self.hc_blocks.extend([norm(ll.Conv1d(args.Cs, args.Cs * 2, 1))])
self.hc_blocks.extend([
norm(ll.HighwayConv1d(args.Cs * 2, args.Cs * 2, 3, dilation=1))
for i in range(2)
])
# (N, Cs*2, Ty) -> (N, n_mags, Ty)
self.hc_blocks.extend([norm(ll.Conv1d(args.Cs * 2, args.n_mags, 1))])
self.hc_blocks.extend([
norm(
ll.Conv1d(args.n_mags,
args.n_mags,
1,
activation_fn=torch.relu)) for i in range(2)
])
self.hc_blocks.extend([norm(ll.Conv1d(args.n_mags, args.n_mags, 1))])