File size: 10,436 Bytes
8c70653
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
# adopted from https://github.com/jik876/hifi-gan/blob/master/models.py
import torch
from torch import nn
from torch.nn import Conv1d, ConvTranspose1d
from torch.nn import functional as F
from torch.nn.utils import remove_weight_norm, weight_norm

from TTS.utils.io import load_fsspec

LRELU_SLOPE = 0.1


def get_padding(k, d):
    return int((k * d - d) / 2)


class ResBlock1(torch.nn.Module):
    """Residual Block Type 1. It has 3 convolutional layers in each convolutiona block.

    Network::

        x -> lrelu -> conv1_1 -> conv1_2 -> conv1_3 -> z -> lrelu -> conv2_1 -> conv2_2 -> conv2_3 -> o -> + -> o
        |--------------------------------------------------------------------------------------------------|


    Args:
        channels (int): number of hidden channels for the convolutional layers.
        kernel_size (int): size of the convolution filter in each layer.
        dilations (list): list of dilation value for each conv layer in a block.
    """

    def __init__(self, channels, kernel_size=3, dilation=(1, 3, 5)):
        super().__init__()
        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]),
                    )
                ),
            ]
        )

        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))
                ),
            ]
        )

    def forward(self, x):
        """
        Args:
            x (Tensor): input tensor.
        Returns:
            Tensor: output tensor.
        Shapes:
            x: [B, C, T]
        """
        for c1, c2 in zip(self.convs1, self.convs2):
            xt = F.leaky_relu(x, LRELU_SLOPE)
            xt = c1(xt)
            xt = F.leaky_relu(xt, LRELU_SLOPE)
            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):
    """Residual Block Type 1. It has 3 convolutional layers in each convolutiona block.

    Network::

        x -> lrelu -> conv1-> -> z -> lrelu -> conv2-> o -> + -> o
        |---------------------------------------------------|


    Args:
        channels (int): number of hidden channels for the convolutional layers.
        kernel_size (int): size of the convolution filter in each layer.
        dilations (list): list of dilation value for each conv layer in a block.
    """

    def __init__(self, channels, kernel_size=3, dilation=(1, 3)):
        super().__init__()
        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]),
                    )
                ),
            ]
        )

    def forward(self, x):
        for c in self.convs:
            xt = F.leaky_relu(x, LRELU_SLOPE)
            xt = c(xt)
            x = xt + x
        return x

    def remove_weight_norm(self):
        for l in self.convs:
            remove_weight_norm(l)


class HifiganGenerator(torch.nn.Module):
    def __init__(
        self,
        in_channels,
        out_channels,
        resblock_type,
        resblock_dilation_sizes,
        resblock_kernel_sizes,
        upsample_kernel_sizes,
        upsample_initial_channel,
        upsample_factors,
        inference_padding=5,
        cond_channels=0,
        conv_pre_weight_norm=True,
        conv_post_weight_norm=True,
        conv_post_bias=True,
    ):
        r"""HiFiGAN Generator with Multi-Receptive Field Fusion (MRF)

        Network:
            x -> lrelu -> upsampling_layer -> resblock1_k1x1 -> z1 -> + -> z_sum / #resblocks -> lrelu -> conv_post_7x1 -> tanh -> o
                                                 ..          -> zI ---|
                                              resblockN_kNx1 -> zN ---'

        Args:
            in_channels (int): number of input tensor channels.
            out_channels (int): number of output tensor channels.
            resblock_type (str): type of the `ResBlock`. '1' or '2'.
            resblock_dilation_sizes (List[List[int]]): list of dilation values in each layer of a `ResBlock`.
            resblock_kernel_sizes (List[int]): list of kernel sizes for each `ResBlock`.
            upsample_kernel_sizes (List[int]): list of kernel sizes for each transposed convolution.
            upsample_initial_channel (int): number of channels for the first upsampling layer. This is divided by 2
                for each consecutive upsampling layer.
            upsample_factors (List[int]): upsampling factors (stride) for each upsampling layer.
            inference_padding (int): constant padding applied to the input at inference time. Defaults to 5.
        """
        super().__init__()
        self.inference_padding = inference_padding
        self.num_kernels = len(resblock_kernel_sizes)
        self.num_upsamples = len(upsample_factors)
        # initial upsampling layers
        self.conv_pre = weight_norm(Conv1d(in_channels, upsample_initial_channel, 7, 1, padding=3))
        resblock = ResBlock1 if resblock_type == "1" else ResBlock2
        # upsampling layers
        self.ups = nn.ModuleList()
        for i, (u, k) in enumerate(zip(upsample_factors, upsample_kernel_sizes)):
            self.ups.append(
                weight_norm(
                    ConvTranspose1d(
                        upsample_initial_channel // (2**i),
                        upsample_initial_channel // (2 ** (i + 1)),
                        k,
                        u,
                        padding=(k - u) // 2,
                    )
                )
            )
        # MRF blocks
        self.resblocks = nn.ModuleList()
        for i in range(len(self.ups)):
            ch = upsample_initial_channel // (2 ** (i + 1))
            for _, (k, d) in enumerate(zip(resblock_kernel_sizes, resblock_dilation_sizes)):
                self.resblocks.append(resblock(ch, k, d))
        # post convolution layer
        self.conv_post = weight_norm(Conv1d(ch, out_channels, 7, 1, padding=3, bias=conv_post_bias))
        if cond_channels > 0:
            self.cond_layer = nn.Conv1d(cond_channels, upsample_initial_channel, 1)

        if not conv_pre_weight_norm:
            remove_weight_norm(self.conv_pre)

        if not conv_post_weight_norm:
            remove_weight_norm(self.conv_post)

    def forward(self, x, g=None):
        """
        Args:
            x (Tensor): feature input tensor.
            g (Tensor): global conditioning input tensor.

        Returns:
            Tensor: output waveform.

        Shapes:
            x: [B, C, T]
            Tensor: [B, 1, T]
        """
        o = self.conv_pre(x)
        if hasattr(self, "cond_layer"):
            o = o + self.cond_layer(g)
        for i in range(self.num_upsamples):
            o = F.leaky_relu(o, LRELU_SLOPE)
            o = self.ups[i](o)
            z_sum = None
            for j in range(self.num_kernels):
                if z_sum is None:
                    z_sum = self.resblocks[i * self.num_kernels + j](o)
                else:
                    z_sum += self.resblocks[i * self.num_kernels + j](o)
            o = z_sum / self.num_kernels
        o = F.leaky_relu(o)
        o = self.conv_post(o)
        o = torch.tanh(o)
        return o

    @torch.no_grad()
    def inference(self, c):
        """
        Args:
            x (Tensor): conditioning input tensor.

        Returns:
            Tensor: output waveform.

        Shapes:
            x: [B, C, T]
            Tensor: [B, 1, T]
        """
        c = c.to(self.conv_pre.weight.device)
        c = torch.nn.functional.pad(c, (self.inference_padding, self.inference_padding), "replicate")
        return self.forward(c)

    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)

    def load_checkpoint(
        self, config, checkpoint_path, eval=False, cache=False
    ):  # pylint: disable=unused-argument, redefined-builtin
        state = load_fsspec(checkpoint_path, map_location=torch.device("cpu"), cache=cache)
        self.load_state_dict(state["model"])
        if eval:
            self.eval()
            assert not self.training
            self.remove_weight_norm()