Upload folder using huggingface_hub

.gitattributes

@@ -1,35 +1,2 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz filter=lfs diff=lfs merge=lfs -text
-*.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
+# Auto detect text files and perform LF normalization
+* text=auto

.gitignore

@@ -0,0 +1,11 @@
+DUMMY1
+DUMMY2
+DUMMY3
+logs
+__pycache__
+.ipynb_checkpoints
+.*.swp
+
+build
+*.c
+monotonic_align/monotonic_align

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2023 Kevin Wang
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -1,13 +1,17 @@
----
-title: VITS2 Chinese
-emoji: 🐨
-colorFrom: pink
-colorTo: yellow
-sdk: gradio
-sdk_version: 3.44.4
-app_file: app.py
-pinned: false
-license: mit
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# VITS2-Chinese 🎶🌟💕
+## 只需上传一段语音素材，程序自动将语音切片、标注、预处理，一键训练
+[English]()
+## 环境配置
+1. 运行
+```
+git clone https://github.com/KevinWang676/VITS2-Chinese.git
+cd VITS2-Chinese
+pip install -r requirements.txt
+```
+2. 运行
+```
+cd monotonic_align
+python setup.py build_ext --inplace
+```
+3. 上传语音文件：请上传一段**中文**、**单说话人**的语音文件，建议为长度大于10分钟的`.wav`文件
+## 语音切片

attentions.py

@@ -0,0 +1,646 @@
+import copy
+import math
+import numpy as np
+import torch
+from torch import nn
+from torch.nn import functional as F
+from torch.nn.utils import remove_weight_norm, weight_norm
+
+import commons
+import modules
+from modules import LayerNorm
+
+
+class Encoder(nn.Module):  # backward compatible vits2 encoder
+    def __init__(
+        self,
+        hidden_channels,
+        filter_channels,
+        n_heads,
+        n_layers,
+        kernel_size=1,
+        p_dropout=0.0,
+        window_size=4,
+        **kwargs
+    ):
+        super().__init__()
+        self.hidden_channels = hidden_channels
+        self.filter_channels = filter_channels
+        self.n_heads = n_heads
+        self.n_layers = n_layers
+        self.kernel_size = kernel_size
+        self.p_dropout = p_dropout
+        self.window_size = window_size
+
+        self.drop = nn.Dropout(p_dropout)
+        self.attn_layers = nn.ModuleList()
+        self.norm_layers_1 = nn.ModuleList()
+        self.ffn_layers = nn.ModuleList()
+        self.norm_layers_2 = nn.ModuleList()
+        # if kwargs has spk_emb_dim, then add a linear layer to project spk_emb_dim to hidden_channels
+        self.cond_layer_idx = self.n_layers
+        if "gin_channels" in kwargs:
+            self.gin_channels = kwargs["gin_channels"]
+            if self.gin_channels != 0:
+                self.spk_emb_linear = nn.Linear(self.gin_channels, self.hidden_channels)
+                # vits2 says 3rd block, so idx is 2 by default
+                self.cond_layer_idx = (
+                    kwargs["cond_layer_idx"] if "cond_layer_idx" in kwargs else 2
+                )
+                assert (
+                    self.cond_layer_idx < self.n_layers
+                ), "cond_layer_idx should be less than n_layers"
+
+        for i in range(self.n_layers):
+            self.attn_layers.append(
+                MultiHeadAttention(
+                    hidden_channels,
+                    hidden_channels,
+                    n_heads,
+                    p_dropout=p_dropout,
+                    window_size=window_size,
+                )
+            )
+            self.norm_layers_1.append(LayerNorm(hidden_channels))
+            self.ffn_layers.append(
+                FFN(
+                    hidden_channels,
+                    hidden_channels,
+                    filter_channels,
+                    kernel_size,
+                    p_dropout=p_dropout,
+                )
+            )
+            self.norm_layers_2.append(LayerNorm(hidden_channels))
+
+    def forward(self, x, x_mask, g=None):
+        attn_mask = x_mask.unsqueeze(2) * x_mask.unsqueeze(-1)
+        x = x * x_mask
+        for i in range(self.n_layers):
+            if i == self.cond_layer_idx and g is not None:
+                g = self.spk_emb_linear(g.transpose(1, 2))
+                g = g.transpose(1, 2)
+                x = x + g
+                x = x * x_mask
+            y = self.attn_layers[i](x, x, attn_mask)
+            y = self.drop(y)
+            x = self.norm_layers_1[i](x + y)
+
+            y = self.ffn_layers[i](x, x_mask)
+            y = self.drop(y)
+            x = self.norm_layers_2[i](x + y)
+        x = x * x_mask
+        return x
+
+
+class Decoder(nn.Module):
+    def __init__(
+        self,
+        hidden_channels,
+        filter_channels,
+        n_heads,
+        n_layers,
+        kernel_size=1,
+        p_dropout=0.0,
+        proximal_bias=False,
+        proximal_init=True,
+        **kwargs
+    ):
+        super().__init__()
+        self.hidden_channels = hidden_channels
+        self.filter_channels = filter_channels
+        self.n_heads = n_heads
+        self.n_layers = n_layers
+        self.kernel_size = kernel_size
+        self.p_dropout = p_dropout
+        self.proximal_bias = proximal_bias
+        self.proximal_init = proximal_init
+
+        self.drop = nn.Dropout(p_dropout)
+        self.self_attn_layers = nn.ModuleList()
+        self.norm_layers_0 = nn.ModuleList()
+        self.encdec_attn_layers = nn.ModuleList()
+        self.norm_layers_1 = nn.ModuleList()
+        self.ffn_layers = nn.ModuleList()
+        self.norm_layers_2 = nn.ModuleList()
+        for i in range(self.n_layers):
+            self.self_attn_layers.append(
+                MultiHeadAttention(
+                    hidden_channels,
+                    hidden_channels,
+                    n_heads,
+                    p_dropout=p_dropout,
+                    proximal_bias=proximal_bias,
+                    proximal_init=proximal_init,
+                )
+            )
+            self.norm_layers_0.append(LayerNorm(hidden_channels))
+            self.encdec_attn_layers.append(
+                MultiHeadAttention(
+                    hidden_channels, hidden_channels, n_heads, p_dropout=p_dropout
+                )
+            )
+            self.norm_layers_1.append(LayerNorm(hidden_channels))
+            self.ffn_layers.append(
+                FFN(
+                    hidden_channels,
+                    hidden_channels,
+                    filter_channels,
+                    kernel_size,
+                    p_dropout=p_dropout,
+                    causal=True,
+                )
+            )
+            self.norm_layers_2.append(LayerNorm(hidden_channels))
+
+    def forward(self, x, x_mask, h, h_mask):
+        """
+        x: decoder input
+        h: encoder output
+        """
+        self_attn_mask = commons.subsequent_mask(x_mask.size(2)).to(
+            device=x.device, dtype=x.dtype
+        )
+        encdec_attn_mask = h_mask.unsqueeze(2) * x_mask.unsqueeze(-1)
+        x = x * x_mask
+        for i in range(self.n_layers):
+            y = self.self_attn_layers[i](x, x, self_attn_mask)
+            y = self.drop(y)
+            x = self.norm_layers_0[i](x + y)
+
+            y = self.encdec_attn_layers[i](x, h, encdec_attn_mask)
+            y = self.drop(y)
+            x = self.norm_layers_1[i](x + y)
+
+            y = self.ffn_layers[i](x, x_mask)
+            y = self.drop(y)
+            x = self.norm_layers_2[i](x + y)
+        x = x * x_mask
+        return x
+
+
+class MultiHeadAttention(nn.Module):
+    def __init__(
+        self,
+        channels,
+        out_channels,
+        n_heads,
+        p_dropout=0.0,
+        window_size=None,
+        heads_share=True,
+        block_length=None,
+        proximal_bias=False,
+        proximal_init=False,
+    ):
+        super().__init__()
+        assert channels % n_heads == 0
+
+        self.channels = channels
+        self.out_channels = out_channels
+        self.n_heads = n_heads
+        self.p_dropout = p_dropout
+        self.window_size = window_size
+        self.heads_share = heads_share
+        self.block_length = block_length
+        self.proximal_bias = proximal_bias
+        self.proximal_init = proximal_init
+        self.attn = None
+
+        self.k_channels = channels // n_heads
+        self.conv_q = nn.Conv1d(channels, channels, 1)
+        self.conv_k = nn.Conv1d(channels, channels, 1)
+        self.conv_v = nn.Conv1d(channels, channels, 1)
+        self.conv_o = nn.Conv1d(channels, out_channels, 1)
+        self.drop = nn.Dropout(p_dropout)
+
+        if window_size is not None:
+            n_heads_rel = 1 if heads_share else n_heads
+            rel_stddev = self.k_channels**-0.5
+            self.emb_rel_k = nn.Parameter(
+                torch.randn(n_heads_rel, window_size * 2 + 1, self.k_channels)
+                * rel_stddev
+            )
+            self.emb_rel_v = nn.Parameter(
+                torch.randn(n_heads_rel, window_size * 2 + 1, self.k_channels)
+                * rel_stddev
+            )
+
+        nn.init.xavier_uniform_(self.conv_q.weight)
+        nn.init.xavier_uniform_(self.conv_k.weight)
+        nn.init.xavier_uniform_(self.conv_v.weight)
+        if proximal_init:
+            with torch.no_grad():
+                self.conv_k.weight.copy_(self.conv_q.weight)
+                self.conv_k.bias.copy_(self.conv_q.bias)
+
+    def forward(self, x, c, attn_mask=None):
+        q = self.conv_q(x)
+        k = self.conv_k(c)
+        v = self.conv_v(c)
+
+        x, self.attn = self.attention(q, k, v, mask=attn_mask)
+
+        x = self.conv_o(x)
+        return x
+
+    def attention(self, query, key, value, mask=None):
+        # reshape [b, d, t] -> [b, n_h, t, d_k]
+        b, d, t_s, t_t = (*key.size(), query.size(2))
+        query = query.view(b, self.n_heads, self.k_channels, t_t).transpose(2, 3)
+        key = key.view(b, self.n_heads, self.k_channels, t_s).transpose(2, 3)
+        value = value.view(b, self.n_heads, self.k_channels, t_s).transpose(2, 3)
+
+        scores = torch.matmul(query / math.sqrt(self.k_channels), key.transpose(-2, -1))
+        if self.window_size is not None:
+            assert (
+                t_s == t_t
+            ), "Relative attention is only available for self-attention."
+            key_relative_embeddings = self._get_relative_embeddings(self.emb_rel_k, t_s)
+            rel_logits = self._matmul_with_relative_keys(
+                query / math.sqrt(self.k_channels), key_relative_embeddings
+            )
+            scores_local = self._relative_position_to_absolute_position(rel_logits)
+            scores = scores + scores_local
+        if self.proximal_bias:
+            assert t_s == t_t, "Proximal bias is only available for self-attention."
+            scores = scores + self._attention_bias_proximal(t_s).to(
+                device=scores.device, dtype=scores.dtype
+            )
+        if mask is not None:
+            scores = scores.masked_fill(mask == 0, -1e4)
+            if self.block_length is not None:
+                assert (
+                    t_s == t_t
+                ), "Local attention is only available for self-attention."
+                block_mask = (
+                    torch.ones_like(scores)
+                    .triu(-self.block_length)
+                    .tril(self.block_length)
+                )
+                scores = scores.masked_fill(block_mask == 0, -1e4)
+        p_attn = F.softmax(scores, dim=-1)  # [b, n_h, t_t, t_s]
+        p_attn = self.drop(p_attn)
+        output = torch.matmul(p_attn, value)
+        if self.window_size is not None:
+            relative_weights = self._absolute_position_to_relative_position(p_attn)
+            value_relative_embeddings = self._get_relative_embeddings(
+                self.emb_rel_v, t_s
+            )
+            output = output + self._matmul_with_relative_values(
+                relative_weights, value_relative_embeddings
+            )
+        output = (
+            output.transpose(2, 3).contiguous().view(b, d, t_t)
+        )  # [b, n_h, t_t, d_k] -> [b, d, t_t]
+        return output, p_attn
+
+    def _matmul_with_relative_values(self, x, y):
+        """
+        x: [b, h, l, m]
+        y: [h or 1, m, d]
+        ret: [b, h, l, d]
+        """
+        ret = torch.matmul(x, y.unsqueeze(0))
+        return ret
+
+    def _matmul_with_relative_keys(self, x, y):
+        """
+        x: [b, h, l, d]
+        y: [h or 1, m, d]
+        ret: [b, h, l, m]
+        """
+        ret = torch.matmul(x, y.unsqueeze(0).transpose(-2, -1))
+        return ret
+
+    def _get_relative_embeddings(self, relative_embeddings, length):
+        max_relative_position = 2 * self.window_size + 1
+        # Pad first before slice to avoid using cond ops.
+        pad_length = max(length - (self.window_size + 1), 0)
+        slice_start_position = max((self.window_size + 1) - length, 0)
+        slice_end_position = slice_start_position + 2 * length - 1
+        if pad_length > 0:
+            padded_relative_embeddings = F.pad(
+                relative_embeddings,
+                commons.convert_pad_shape([[0, 0], [pad_length, pad_length], [0, 0]]),
+            )
+        else:
+            padded_relative_embeddings = relative_embeddings
+        used_relative_embeddings = padded_relative_embeddings[
+            :, slice_start_position:slice_end_position
+        ]
+        return used_relative_embeddings
+
+    def _relative_position_to_absolute_position(self, x):
+        """
+        x: [b, h, l, 2*l-1]
+        ret: [b, h, l, l]
+        """
+        batch, heads, length, _ = x.size()
+        # Concat columns of pad to shift from relative to absolute indexing.
+        x = F.pad(x, commons.convert_pad_shape([[0, 0], [0, 0], [0, 0], [0, 1]]))
+
+        # Concat extra elements so to add up to shape (len+1, 2*len-1).
+        x_flat = x.view([batch, heads, length * 2 * length])
+        x_flat = F.pad(
+            x_flat, commons.convert_pad_shape([[0, 0], [0, 0], [0, length - 1]])
+        )
+
+        # Reshape and slice out the padded elements.
+        x_final = x_flat.view([batch, heads, length + 1, 2 * length - 1])[
+            :, :, :length, length - 1 :
+        ]
+        return x_final
+
+    def _absolute_position_to_relative_position(self, x):
+        """
+        x: [b, h, l, l]
+        ret: [b, h, l, 2*l-1]
+        """
+        batch, heads, length, _ = x.size()
+        # padd along column
+        x = F.pad(
+            x, commons.convert_pad_shape([[0, 0], [0, 0], [0, 0], [0, length - 1]])
+        )
+        x_flat = x.view([batch, heads, length**2 + length * (length - 1)])
+        # add 0's in the beginning that will skew the elements after reshape
+        x_flat = F.pad(x_flat, commons.convert_pad_shape([[0, 0], [0, 0], [length, 0]]))
+        x_final = x_flat.view([batch, heads, length, 2 * length])[:, :, :, 1:]
+        return x_final
+
+    def _attention_bias_proximal(self, length):
+        """Bias for self-attention to encourage attention to close positions.
+        Args:
+          length: an integer scalar.
+        Returns:
+          a Tensor with shape [1, 1, length, length]
+        """
+        r = torch.arange(length, dtype=torch.float32)
+        diff = torch.unsqueeze(r, 0) - torch.unsqueeze(r, 1)
+        return torch.unsqueeze(torch.unsqueeze(-torch.log1p(torch.abs(diff)), 0), 0)
+
+
+class FFN(nn.Module):
+    def __init__(
+        self,
+        in_channels,
+        out_channels,
+        filter_channels,
+        kernel_size,
+        p_dropout=0.0,
+        activation=None,
+        causal=False,
+    ):
+        super().__init__()
+        self.in_channels = in_channels
+        self.out_channels = out_channels
+        self.filter_channels = filter_channels
+        self.kernel_size = kernel_size
+        self.p_dropout = p_dropout
+        self.activation = activation
+        self.causal = causal
+
+        if causal:
+            self.padding = self._causal_padding
+        else:
+            self.padding = self._same_padding
+
+        self.conv_1 = nn.Conv1d(in_channels, filter_channels, kernel_size)
+        self.conv_2 = nn.Conv1d(filter_channels, out_channels, kernel_size)
+        self.drop = nn.Dropout(p_dropout)
+
+    def forward(self, x, x_mask):
+        x = self.conv_1(self.padding(x * x_mask))
+        if self.activation == "gelu":
+            x = x * torch.sigmoid(1.702 * x)
+        else:
+            x = torch.relu(x)
+        x = self.drop(x)
+        x = self.conv_2(self.padding(x * x_mask))
+        return x * x_mask
+
+    def _causal_padding(self, x):
+        if self.kernel_size == 1:
+            return x
+        pad_l = self.kernel_size - 1
+        pad_r = 0
+        padding = [[0, 0], [0, 0], [pad_l, pad_r]]
+        x = F.pad(x, commons.convert_pad_shape(padding))
+        return x
+
+    def _same_padding(self, x):
+        if self.kernel_size == 1:
+            return x
+        pad_l = (self.kernel_size - 1) // 2
+        pad_r = self.kernel_size // 2
+        padding = [[0, 0], [0, 0], [pad_l, pad_r]]
+        x = F.pad(x, commons.convert_pad_shape(padding))
+        return x
+
+
+class Depthwise_Separable_Conv1D(nn.Module):
+    def __init__(
+        self,
+        in_channels,
+        out_channels,
+        kernel_size,
+        stride=1,
+        padding=0,
+        dilation=1,
+        bias=True,
+        padding_mode="zeros",  # TODO: refine this type
+        device=None,
+        dtype=None,
+    ):
+        super().__init__()
+        self.depth_conv = nn.Conv1d(
+            in_channels=in_channels,
+            out_channels=in_channels,
+            kernel_size=kernel_size,
+            groups=in_channels,
+            stride=stride,
+            padding=padding,
+            dilation=dilation,
+            bias=bias,
+            padding_mode=padding_mode,
+            device=device,
+            dtype=dtype,
+        )
+        self.point_conv = nn.Conv1d(
+            in_channels=in_channels,
+            out_channels=out_channels,
+            kernel_size=1,
+            bias=bias,
+            device=device,
+            dtype=dtype,
+        )
+
+    def forward(self, input):
+        return self.point_conv(self.depth_conv(input))
+
+    def weight_norm(self):
+        self.depth_conv = weight_norm(self.depth_conv, name="weight")
+        self.point_conv = weight_norm(self.point_conv, name="weight")
+
+    def remove_weight_norm(self):
+        self.depth_conv = remove_weight_norm(self.depth_conv, name="weight")
+        self.point_conv = remove_weight_norm(self.point_conv, name="weight")
+
+
+class Depthwise_Separable_TransposeConv1D(nn.Module):
+    def __init__(
+        self,
+        in_channels,
+        out_channels,
+        kernel_size,
+        stride=1,
+        padding=0,
+        output_padding=0,
+        bias=True,
+        dilation=1,
+        padding_mode="zeros",  # TODO: refine this type
+        device=None,
+        dtype=None,
+    ):
+        super().__init__()
+        self.depth_conv = nn.ConvTranspose1d(
+            in_channels=in_channels,
+            out_channels=in_channels,
+            kernel_size=kernel_size,
+            groups=in_channels,
+            stride=stride,
+            output_padding=output_padding,
+            padding=padding,
+            dilation=dilation,
+            bias=bias,
+            padding_mode=padding_mode,
+            device=device,
+            dtype=dtype,
+        )
+        self.point_conv = nn.Conv1d(
+            in_channels=in_channels,
+            out_channels=out_channels,
+            kernel_size=1,
+            bias=bias,
+            device=device,
+            dtype=dtype,
+        )
+
+    def forward(self, input):
+        return self.point_conv(self.depth_conv(input))
+
+    def weight_norm(self):
+        self.depth_conv = weight_norm(self.depth_conv, name="weight")
+        self.point_conv = weight_norm(self.point_conv, name="weight")
+
+    def remove_weight_norm(self):
+        remove_weight_norm(self.depth_conv, name="weight")
+        remove_weight_norm(self.point_conv, name="weight")
+
+
+def weight_norm_modules(module, name="weight", dim=0):
+    if isinstance(module, Depthwise_Separable_Conv1D) or isinstance(
+        module, Depthwise_Separable_TransposeConv1D
+    ):
+        module.weight_norm()
+        return module
+    else:
+        return weight_norm(module, name, dim)
+
+
+def remove_weight_norm_modules(module, name="weight"):
+    if isinstance(module, Depthwise_Separable_Conv1D) or isinstance(
+        module, Depthwise_Separable_TransposeConv1D
+    ):
+        module.remove_weight_norm()
+    else:
+        remove_weight_norm(module, name)
+
+
+class FFT(nn.Module):
+    def __init__(
+        self,
+        hidden_channels,
+        filter_channels,
+        n_heads,
+        n_layers=1,
+        kernel_size=1,
+        p_dropout=0.0,
+        proximal_bias=False,
+        proximal_init=True,
+        isflow=False,
+        **kwargs
+    ):
+        super().__init__()
+        self.hidden_channels = hidden_channels
+        self.filter_channels = filter_channels
+        self.n_heads = n_heads
+        self.n_layers = n_layers
+        self.kernel_size = kernel_size
+        self.p_dropout = p_dropout
+        self.proximal_bias = proximal_bias
+        self.proximal_init = proximal_init
+        if isflow and "gin_channels" in kwargs and kwargs["gin_channels"] > 0:
+            cond_layer = torch.nn.Conv1d(
+                kwargs["gin_channels"], 2 * hidden_channels * n_layers, 1
+            )
+            self.cond_pre = torch.nn.Conv1d(hidden_channels, 2 * hidden_channels, 1)
+            self.cond_layer = weight_norm_modules(cond_layer, name="weight")
+            self.gin_channels = kwargs["gin_channels"]
+        self.drop = nn.Dropout(p_dropout)
+        self.self_attn_layers = nn.ModuleList()
+        self.norm_layers_0 = nn.ModuleList()
+        self.ffn_layers = nn.ModuleList()
+        self.norm_layers_1 = nn.ModuleList()
+        for i in range(self.n_layers):
+            self.self_attn_layers.append(
+                MultiHeadAttention(
+                    hidden_channels,
+                    hidden_channels,
+                    n_heads,
+                    p_dropout=p_dropout,
+                    proximal_bias=proximal_bias,
+                    proximal_init=proximal_init,
+                )
+            )
+            self.norm_layers_0.append(LayerNorm(hidden_channels))
+            self.ffn_layers.append(
+                FFN(
+                    hidden_channels,
+                    hidden_channels,
+                    filter_channels,
+                    kernel_size,
+                    p_dropout=p_dropout,
+                    causal=True,
+                )
+            )
+            self.norm_layers_1.append(LayerNorm(hidden_channels))
+
+    def forward(self, x, x_mask, g=None):
+        """
+        x: decoder input
+        h: encoder output
+        """
+        if g is not None:
+            g = self.cond_layer(g)
+
+        self_attn_mask = commons.subsequent_mask(x_mask.size(2)).to(
+            device=x.device, dtype=x.dtype
+        )
+        x = x * x_mask
+        for i in range(self.n_layers):
+            if g is not None:
+                x = self.cond_pre(x)
+                cond_offset = i * 2 * self.hidden_channels
+                g_l = g[:, cond_offset : cond_offset + 2 * self.hidden_channels, :]
+                x = commons.fused_add_tanh_sigmoid_multiply(
+                    x, g_l, torch.IntTensor([self.hidden_channels])
+                )
+            y = self.self_attn_layers[i](x, x, self_attn_mask)
+            y = self.drop(y)
+            x = self.norm_layers_0[i](x + y)
+
+            y = self.ffn_layers[i](x, x_mask)
+            y = self.drop(y)
+            x = self.norm_layers_1[i](x + y)
+        x = x * x_mask
+        return x

commons.py

@@ -0,0 +1,163 @@
+import math
+import numpy as np
+import torch
+from torch import nn
+from torch.nn import functional as F
+
+
+def init_weights(m, mean=0.0, std=0.01):
+    classname = m.__class__.__name__
+    if classname.find("Conv") != -1:
+        m.weight.data.normal_(mean, std)
+
+
+def get_padding(kernel_size, dilation=1):
+    return int((kernel_size * dilation - dilation) / 2)
+
+
+def convert_pad_shape(pad_shape):
+    l = pad_shape[::-1]
+    pad_shape = [item for sublist in l for item in sublist]
+    return pad_shape
+
+
+def intersperse(lst, item):
+    result = [item] * (len(lst) * 2 + 1)
+    result[1::2] = lst
+    return result
+
+
+def kl_divergence(m_p, logs_p, m_q, logs_q):
+    """KL(P||Q)"""
+    kl = (logs_q - logs_p) - 0.5
+    kl += (
+        0.5 * (torch.exp(2.0 * logs_p) + ((m_p - m_q) ** 2)) * torch.exp(-2.0 * logs_q)
+    )
+    return kl
+
+
+def rand_gumbel(shape):
+    """Sample from the Gumbel distribution, protect from overflows."""
+    uniform_samples = torch.rand(shape) * 0.99998 + 0.00001
+    return -torch.log(-torch.log(uniform_samples))
+
+
+def rand_gumbel_like(x):
+    g = rand_gumbel(x.size()).to(dtype=x.dtype, device=x.device)
+    return g
+
+
+def slice_segments(x, ids_str, segment_size=4):
+    ret = torch.zeros_like(x[:, :, :segment_size])
+    for i in range(x.size(0)):
+        idx_str = ids_str[i]
+        idx_end = idx_str + segment_size
+        ret[i] = x[i, :, idx_str:idx_end]
+    return ret
+
+
+def rand_slice_segments(x, x_lengths=None, segment_size=4):
+    b, d, t = x.size()
+    if x_lengths is None:
+        x_lengths = t
+    ids_str_max = x_lengths - segment_size + 1
+    ids_str = (torch.rand([b]).to(device=x.device) * ids_str_max).to(dtype=torch.long)
+    ret = slice_segments(x, ids_str, segment_size)
+    return ret, ids_str
+
+
+def get_timing_signal_1d(length, channels, min_timescale=1.0, max_timescale=1.0e4):
+    position = torch.arange(length, dtype=torch.float)
+    num_timescales = channels // 2
+    log_timescale_increment = math.log(float(max_timescale) / float(min_timescale)) / (
+        num_timescales - 1
+    )
+    inv_timescales = min_timescale * torch.exp(
+        torch.arange(num_timescales, dtype=torch.float) * -log_timescale_increment
+    )
+    scaled_time = position.unsqueeze(0) * inv_timescales.unsqueeze(1)
+    signal = torch.cat([torch.sin(scaled_time), torch.cos(scaled_time)], 0)
+    signal = F.pad(signal, [0, 0, 0, channels % 2])
+    signal = signal.view(1, channels, length)
+    return signal
+
+
+def add_timing_signal_1d(x, min_timescale=1.0, max_timescale=1.0e4):
+    b, channels, length = x.size()
+    signal = get_timing_signal_1d(length, channels, min_timescale, max_timescale)
+    return x + signal.to(dtype=x.dtype, device=x.device)
+
+
+def cat_timing_signal_1d(x, min_timescale=1.0, max_timescale=1.0e4, axis=1):
+    b, channels, length = x.size()
+    signal = get_timing_signal_1d(length, channels, min_timescale, max_timescale)
+    return torch.cat([x, signal.to(dtype=x.dtype, device=x.device)], axis)
+
+
+def subsequent_mask(length):
+    mask = torch.tril(torch.ones(length, length)).unsqueeze(0).unsqueeze(0)
+    return mask
+
+
+@torch.jit.script
+def fused_add_tanh_sigmoid_multiply(input_a, input_b, n_channels):
+    n_channels_int = n_channels[0]
+    in_act = input_a + input_b
+    t_act = torch.tanh(in_act[:, :n_channels_int, :])
+    s_act = torch.sigmoid(in_act[:, n_channels_int:, :])
+    acts = t_act * s_act
+    return acts
+
+
+def convert_pad_shape(pad_shape):
+    l = pad_shape[::-1]
+    pad_shape = [item for sublist in l for item in sublist]
+    return pad_shape
+
+
+def shift_1d(x):
+    x = F.pad(x, convert_pad_shape([[0, 0], [0, 0], [1, 0]]))[:, :, :-1]
+    return x
+
+
+def sequence_mask(length, max_length=None):
+    if max_length is None:
+        max_length = length.max()
+    x = torch.arange(max_length, dtype=length.dtype, device=length.device)
+    return x.unsqueeze(0) < length.unsqueeze(1)
+
+
+def generate_path(duration, mask):
+    """
+    duration: [b, 1, t_x]
+    mask: [b, 1, t_y, t_x]
+    """
+    device = duration.device
+
+    b, _, t_y, t_x = mask.shape
+    cum_duration = torch.cumsum(duration, -1)
+
+    cum_duration_flat = cum_duration.view(b * t_x)
+    path = sequence_mask(cum_duration_flat, t_y).to(mask.dtype)
+    path = path.view(b, t_x, t_y)
+    path = path - F.pad(path, convert_pad_shape([[0, 0], [1, 0], [0, 0]]))[:, :-1]
+    path = path.unsqueeze(1).transpose(2, 3) * mask
+    return path
+
+
+def clip_grad_value_(parameters, clip_value, norm_type=2):
+    if isinstance(parameters, torch.Tensor):
+        parameters = [parameters]
+    parameters = list(filter(lambda p: p.grad is not None, parameters))
+    norm_type = float(norm_type)
+    if clip_value is not None:
+        clip_value = float(clip_value)
+
+    total_norm = 0
+    for p in parameters:
+        param_norm = p.grad.data.norm(norm_type)
+        total_norm += param_norm.item() ** norm_type
+        if clip_value is not None:
+            p.grad.data.clamp_(min=-clip_value, max=clip_value)
+    total_norm = total_norm ** (1.0 / norm_type)
+    return total_norm

configs/config.json

@@ -0,0 +1,60 @@
+{
+    "train": {
+      "log_interval": 10,
+      "eval_interval": 100,
+      "seed": 1234,
+      "epochs": 1000,
+      "learning_rate": 0.0002,
+      "betas": [0.8, 0.99],
+      "eps": 1e-9,
+      "batch_size": 12,
+      "fp16_run": false,
+      "lr_decay": 0.999875,
+      "segment_size": 8192,
+      "init_lr_ratio": 1,
+      "warmup_epochs": 0,
+      "c_mel": 45,
+      "c_kl": 1.0
+    },
+    "data": {
+      "use_mel_posterior_encoder": true,
+      "training_files":"filelists/final_annotation_train.txt",
+      "validation_files":"filelists/final_annotation_val.txt",
+      "text_cleaners":["chinese_cleaners"],
+      "max_wav_value": 32768.0,
+      "sampling_rate": 22050,
+      "filter_length": 1024,
+      "hop_length": 256,
+      "win_length": 1024,
+      "n_mel_channels": 80,
+      "mel_fmin": 0.0,
+      "mel_fmax": null,
+      "add_blank": false,
+      "n_speakers": 0,
+      "cleaned_text": true
+    },
+    "model": {
+      "use_mel_posterior_encoder": true,
+      "use_transformer_flows": true,
+      "transformer_flow_type": "pre_conv",
+      "use_spk_conditioned_encoder": false,
+      "use_noise_scaled_mas": true,
+      "use_duration_discriminator": true,
+      "inter_channels": 192,
+      "hidden_channels": 192,
+      "filter_channels": 768,
+      "n_heads": 2,
+      "n_layers": 6,
+      "kernel_size": 3,
+      "p_dropout": 0.1,
+      "resblock": "1",
+      "resblock_kernel_sizes": [3,7,11],
+      "resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
+      "upsample_rates": [8,8,2,2],
+      "upsample_initial_channel": 512,
+      "upsample_kernel_sizes": [16,16,4,4],
+      "n_layers_q": 3,
+      "use_spectral_norm": false
+    }
+  }
+  

configs/finetune_speaker.json

@@ -0,0 +1,55 @@
+{
+  "train": {
+    "log_interval": 200,
+    "eval_interval": 1000,
+    "seed": 1234,
+    "epochs": 10000,
+    "learning_rate": 2e-4,
+    "betas": [0.8, 0.99],
+    "eps": 1e-9,
+    "batch_size": 64,
+    "fp16_run": true,
+    "lr_decay": 0.999875,
+    "segment_size": 8192,
+    "init_lr_ratio": 1,
+    "warmup_epochs": 0,
+    "c_mel": 45,
+    "c_kl": 1.0
+  },
+  "data": {
+    "training_files":"filelists/uma_genshin_genshinjp_bh3_train.txt.cleaned",
+    "validation_files":"filelists/uma_genshin_genshinjp_bh3_val.txt.cleaned",
+    "text_cleaners":["chinese_cleaners"],
+    "max_wav_value": 32768.0,
+    "sampling_rate": 22050,
+    "filter_length": 1024,
+    "hop_length": 256,
+    "win_length": 1024,
+    "n_mel_channels": 80,
+    "mel_fmin": 0.0,
+    "mel_fmax": null,
+    "add_blank": true,
+    "n_speakers": 804,
+    "cleaned_text": true
+  },
+  "model": {
+    "inter_channels": 192,
+    "hidden_channels": 192,
+    "filter_channels": 768,
+    "n_heads": 2,
+    "n_layers": 6,
+    "kernel_size": 3,
+    "p_dropout": 0.1,
+    "resblock": "1",
+    "resblock_kernel_sizes": [3,7,11],
+    "resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
+    "upsample_rates": [8,8,2,2],
+    "upsample_initial_channel": 512,
+    "upsample_kernel_sizes": [16,16,4,4],
+    "n_layers_q": 3,
+    "use_spectral_norm": false,
+    "gin_channels": 256
+  },
+  "speakers": ["\u7279\u522b\u5468", "\u65e0\u58f0\u94c3\u9e7f", "\u4e1c\u6d77\u5e1d\u7687\uff08\u5e1d\u5b9d\uff0c\u5e1d\u738b\uff09", "\u4e38\u5584\u65af\u57fa", "\u5bcc\u58eb\u5947\u8ff9", "\u5c0f\u6817\u5e3d", "\u9ec4\u91d1\u8239", "\u4f0f\u7279\u52a0", "\u5927\u548c\u8d64\u9aa5", "\u5927\u6811\u5feb\u8f66", "\u8349\u4e0a\u98de", "\u83f1\u4e9a\u9a6c\u900a", "\u76ee\u767d\u9ea6\u6606", "\u795e\u9e70", "\u597d\u6b4c\u5267", "\u6210\u7530\u767d\u4ec1", "\u9c81\u9053\u592b\u8c61\u5f81\uff08\u7687\u5e1d\uff09", "\u6c14\u69fd", "\u7231\u4e3d\u6570\u7801", "\u661f\u4e91\u5929\u7a7a", "\u7389\u85fb\u5341\u5b57", "\u7f8e\u5999\u59ff\u52bf", "\u7435\u7436\u6668\u5149", "\u6469\u8036\u91cd\u70ae", "\u66fc\u57ce\u8336\u5ea7", "\u7f8e\u6d66\u6ce2\u65c1", "\u76ee\u767d\u8d56\u6069", "\u83f1\u66d9", "\u96ea\u4e2d\u7f8e\u4eba", "\u7c73\u6d74", "\u827e\u5c3c\u65af\u98ce\u795e", "\u7231\u4e3d\u901f\u5b50\uff08\u7231\u4e3d\u5feb\u5b50\uff09", "\u7231\u6155\u7ec7\u59ec", "\u7a3b\u8377\u4e00", "\u80dc\u5229\u5956\u5238", "\u7a7a\u4e2d\u795e\u5bab", "\u8363\u8fdb\u95ea\u8000", "\u771f\u673a\u4f36", "\u5ddd\u4e0a\u516c\u4e3b", "\u9ec4\u91d1\u57ce\uff08\u9ec4\u91d1\u57ce\u5e02\uff09", "\u6a31\u82b1\u8fdb\u738b", "\u91c7\u73e0", "\u65b0\u5149\u98ce", "\u4e1c\u5546\u53d8\u9769", "\u8d85\u7ea7\u5c0f\u6d77\u6e7e", "\u9192\u76ee\u98de\u9e70\uff08\u5bc4\u5bc4\u5b50\uff09", "\u8352\u6f20\u82f1\u96c4", "\u4e1c\u701b\u4f50\u6566", "\u4e2d\u5c71\u5e86\u5178", "\u6210\u7530\u5927\u8fdb", "\u897f\u91ce\u82b1", "\u6625\u4e3d\uff08\u4e4c\u62c9\u62c9\uff09", "\u9752\u7af9\u56de\u5fc6", "\u5fae\u5149\u98de\u9a79", "\u7f8e\u4e3d\u5468\u65e5", "\u5f85\u517c\u798f\u6765", "mr cb\uff08cb\u5148\u751f\uff09", "\u540d\u5c06\u6012\u6d9b\uff08\u540d\u5c06\u6237\u4ec1\uff09", "\u76ee\u767d\u591a\u4f2f", "\u4f18\u79c0\u7d20\u8d28", "\u5e1d\u738b\u5149\u8f89", "\u5f85\u517c\u8bd7\u6b4c\u5267", "\u751f\u91ce\u72c4\u675c\u65af", "\u76ee\u767d\u5584\u4fe1", "\u5927\u62d3\u592a\u9633\u795e", "\u53cc\u6da1\u8f6e\uff08\u4e24\u7acb\u76f4\uff0c\u4e24\u55b7\u5c04\uff0c\u4e8c\u9505\u5934\uff0c\u9006\u55b7\u5c04\uff09", "\u91cc\u89c1\u5149\u94bb\uff08\u8428\u6258\u8bfa\u91d1\u521a\u77f3\uff09", "\u5317\u90e8\u7384\u9a79", "\u6a31\u82b1\u5343\u4ee3\u738b", "\u5929\u72fc\u661f\u8c61\u5f81", "\u76ee\u767d\u963f\u5c14\u4e39", "\u516b\u91cd\u65e0\u654c", "\u9e64\u4e38\u521a\u5fd7", "\u76ee\u767d\u5149\u660e", "\u6210\u7530\u62dc\u4ec1\uff08\u6210\u7530\u8def\uff09", "\u4e5f\u6587\u6444\u8f89", "\u5c0f\u6797\u5386\u5947", "\u5317\u6e2f\u706b\u5c71", "\u5947\u9510\u9a8f", "\u82e6\u6da9\u7cd6\u971c", "\u5c0f\u5c0f\u8695\u8327", "\u9a8f\u5ddd\u624b\u7eb2\uff08\u7eff\u5e3d\u6076\u9b54\uff09", "\u79cb\u5ddd\u5f25\u751f\uff08\u5c0f\u5c0f\u7406\u4e8b\u957f\uff09", "\u4e59\u540d\u53f2\u60a6\u5b50\uff08\u4e59\u540d\u8bb0\u8005\uff09", "\u6850\u751f\u9662\u8475", "\u5b89\u5fc3\u6cfd\u523a\u523a\u7f8e", "\u6a2b\u672c\u7406\u5b50", "\u795e\u91cc\u7eeb\u534e\uff08\u9f9f\u9f9f\uff09", "\u7434", "\u7a7a\uff08\u7a7a\u54e5\uff09", "\u4e3d\u838e", "\u8367\uff08\u8367\u59b9\uff09", "\u82ad\u82ad\u62c9", "\u51ef\u4e9a", "\u8fea\u5362\u514b", "\u96f7\u6cfd", "\u5b89\u67cf", "\u6e29\u8fea", "\u9999\u83f1", "\u5317\u6597", "\u884c\u79cb", "\u9b48", "\u51dd\u5149", "\u53ef\u8389", "\u949f\u79bb", "\u83f2\u8c22\u5c14\uff08\u7687\u5973\uff09", "\u73ed\u5c3c\u7279", "\u8fbe\u8fbe\u5229\u4e9a\uff08\u516c\u5b50\uff09", "\u8bfa\u827e\u5c14\uff08\u5973\u4ec6\uff09", "\u4e03\u4e03", "\u91cd\u4e91", "\u7518\u96e8\uff08\u6930\u7f8a\uff09", "\u963f\u8d1d\u591a", "\u8fea\u5965\u5a1c\uff08\u732b\u732b\uff09", "\u83ab\u5a1c", "\u523b\u6674", "\u7802\u7cd6", "\u8f9b\u7131", "\u7f57\u838e\u8389\u4e9a", "\u80e1\u6843", "\u67ab\u539f\u4e07\u53f6\uff08\u4e07\u53f6\uff09", "\u70df\u7eef", "\u5bb5\u5bab", "\u6258\u9a6c", "\u4f18\u83c8", "\u96f7\u7535\u5c06\u519b\uff08\u96f7\u795e\uff09", "\u65e9\u67da", "\u73ca\u745a\u5bab\u5fc3\u6d77\uff08\u5fc3\u6d77\uff0c\u6263\u6263\u7c73\uff09", "\u4e94\u90ce", "\u4e5d\u6761\u88df\u7f57", "\u8352\u6cf7\u4e00\u6597\uff08\u4e00\u6597\uff09", "\u57c3\u6d1b\u4f0a", "\u7533\u9e64", "\u516b\u91cd\u795e\u5b50\uff08\u795e\u5b50\uff09", "\u795e\u91cc\u7eeb\u4eba\uff08\u7eeb\u4eba\uff09", "\u591c\u5170", "\u4e45\u5c90\u5fcd", "\u9e7f\u91ce\u82d1\u5e73\u85cf", "\u63d0\u7eb3\u91cc", "\u67ef\u83b1", "\u591a\u8389", "\u4e91\u5807", "\u7eb3\u897f\u59b2\uff08\u8349\u795e\uff09", "\u6df1\u6e0a\u4f7f\u5f92", "\u59ae\u9732", "\u8d5b\u8bfa", "\u503a\u52a1\u5904\u7406\u4eba", "\u574e\u8482\u4e1d", "\u771f\u5f13\u5feb\u8f66", "\u79cb\u4eba", "\u671b\u65cf", "\u827e\u5c14\u83f2", "\u827e\u8389\u4e1d", "\u827e\u4f26", "\u963f\u6d1b\u74e6", "\u5929\u91ce", "\u5929\u76ee\u5341\u4e94", "\u611a\u4eba\u4f17-\u5b89\u5fb7\u70c8", "\u5b89\u987a", "\u5b89\u897f", "\u8475", "\u9752\u6728", "\u8352\u5ddd\u5e78\u6b21", "\u8352\u8c37", "\u6709\u6cfd", "\u6d45\u5ddd", "\u9ebb\u7f8e", "\u51dd\u5149\u52a9\u624b", "\u963f\u6258", "\u7afa\u5b50", "\u767e\u8bc6", "\u767e\u95fb", "\u767e\u6653", "\u767d\u672f", "\u8d1d\u96c5\u7279\u4e3d\u5947", "\u4e3d\u5854", "\u5931\u843d\u8ff7\u8fed", "\u7f2d\u4e71\u661f\u68d8", "\u4f0a\u7538", "\u4f0f\u7279\u52a0\u5973\u5b69", "\u72c2\u70ed\u84dd\u8c03", "\u8389\u8389\u5a05", "\u841d\u838e\u8389\u5a05", "\u516b\u91cd\u6a31", "\u516b\u91cd\u971e", "\u5361\u83b2", "\u7b2c\u516d\u591c\u60f3\u66f2", "\u5361\u841d\u5c14", "\u59ec\u5b50", "\u6781\u5730\u6218\u5203", "\u5e03\u6d1b\u59ae\u5a05", "\u6b21\u751f\u94f6\u7ffc", "\u7406\u4e4b\u5f8b\u8005%26\u5e0c\u513f", "\u7406\u4e4b\u5f8b\u8005", "\u8ff7\u57ce\u9a87\u5154", "\u5e0c\u513f", "\u9b47\u591c\u661f\u6e0a", "\u9ed1\u5e0c\u513f", "\u5e15\u6735\u83f2\u8389\u4e1d", "\u4e0d\u706d\u661f\u951a", "\u5929\u5143\u9a91\u82f1", "\u5e7d\u5170\u9edb\u5c14", "\u6d3e\u8499bh3", "\u7231\u9171", "\u7eef\u7389\u4e38", "\u5fb7\u4e3d\u838e", "\u6708\u4e0b\u521d\u62e5", "\u6714\u591c\u89c2\u661f", "\u66ae\u5149\u9a91\u58eb", "\u683c\u857e\u4fee", "\u7559\u4e91\u501f\u98ce\u771f\u541b", "\u6885\u6bd4\u4e4c\u65af", "\u4eff\u72b9\u5927", "\u514b\u83b1\u56e0", "\u5723\u5251\u5e7d\u5170\u9edb\u5c14", "\u5996\u7cbe\u7231\u8389", "\u7279\u65af\u62c9zero", "\u82cd\u7384", "\u82e5\u6c34", "\u897f\u7433", "\u6234\u56e0\u65af\u96f7\u5e03", "\u8d1d\u62c9", "\u8d64\u9e22", "\u9547\u9b42\u6b4c", "\u6e21\u9e26", "\u4eba\u4e4b\u5f8b\u8005", "\u7231\u8389\u5e0c\u96c5", "\u5929\u7a79\u6e38\u4fa0", "\u742a\u4e9a\u5a1c", "\u7a7a\u4e4b\u5f8b\u8005", "\u85aa\u708e\u4e4b\u5f8b\u8005", "\u4e91\u58a8\u4e39\u5fc3", "\u7b26\u534e", "\u8bc6\u4e4b\u5f8b\u8005", "\u7279\u74e6\u6797", "\u7ef4\u5c14\u8587", "\u82bd\u8863", "\u96f7\u4e4b\u5f8b\u8005", "\u65ad\u7f6a\u5f71\u821e", "\u963f\u6ce2\u5c3c\u4e9a", "\u698e\u672c", "\u5384\u5c3c\u65af\u7279", "\u6076\u9f99", "\u8303\u4e8c\u7237", "\u6cd5\u62c9", "\u611a\u4eba\u4f17\u58eb\u5175", "\u611a\u4eba\u4f17\u58eb\u5175a", "\u611a\u4eba\u4f17\u58eb\u5175b", "\u611a\u4eba\u4f17\u58eb\u5175c", "\u611a\u4eba\u4f17a", "\u611a\u4eba\u4f17b", "\u98de\u98de", "\u83f2\u5229\u514b\u65af", "\u5973\u6027\u8ddf\u968f\u8005", "\u9022\u5ca9", "\u6446\u6e21\u4eba", "\u72c2\u8e81\u7684\u7537\u4eba", "\u5965\u5179", "\u8299\u841d\u62c9", "\u8ddf\u968f\u8005", "\u871c\u6c41\u751f\u7269", "\u9ec4\u9ebb\u5b50", "\u6e0a\u4e0a", "\u85e4\u6728", "\u6df1\u89c1", "\u798f\u672c", "\u8299\u84c9", "\u53e4\u6cfd", "\u53e4\u7530", "\u53e4\u5c71", "\u53e4\u8c37\u6607", "\u5085\u4e09\u513f", "\u9ad8\u8001\u516d", "\u77ff\u5de5\u5192", "\u5143\u592a", "\u5fb7\u5b89\u516c", "\u8302\u624d\u516c", "\u6770\u62c9\u5fb7", "\u845b\u7f57\u4e3d", "\u91d1\u5ffd\u5f8b", "\u516c\u4fca", "\u9505\u5df4", "\u6b4c\u5fb7", "\u963f\u8c6a", "\u72d7\u4e09\u513f", "\u845b\u745e\u4e1d", "\u82e5\u5fc3", "\u963f\u5c71\u5a46", "\u602a\u9e1f", "\u5e7f\u7af9", "\u89c2\u6d77", "\u5173\u5b8f", "\u871c\u6c41\u536b\u5175", "\u5b88\u536b1", "\u50b2\u6162\u7684\u5b88\u536b", "\u5bb3\u6015\u7684\u5b88\u536b", "\u8d35\u5b89", "\u76d6\u4f0a", "\u963f\u521b", "\u54c8\u592b\u4e39", "\u65e5\u8bed\u963f\u8d1d\u591a\uff08\u91ce\u5c9b\u5065\u513f\uff09", "\u65e5\u8bed\u57c3\u6d1b\u4f0a\uff08\u9ad8\u57a3\u5f69\u9633\uff09", "\u65e5\u8bed\u5b89\u67cf\uff08\u77f3\u89c1\u821e\u83dc\u9999\uff09", "\u65e5\u8bed\u795e\u91cc\u7eeb\u534e\uff08\u65e9\u89c1\u6c99\u7ec7\uff09", "\u65e5\u8bed\u795e\u91cc\u7eeb\u4eba\uff08\u77f3\u7530\u5f70\uff09", "\u65e5\u8bed\u767d\u672f\uff08\u6e38\u4f50\u6d69\u4e8c\uff09", "\u65e5\u8bed\u82ad\u82ad\u62c9\uff08\u9b3c\u5934\u660e\u91cc\uff09", "\u65e5\u8bed\u5317\u6597\uff08\u5c0f\u6e05\u6c34\u4e9a\u7f8e\uff09", "\u65e5\u8bed\u73ed\u5c3c\u7279\uff08\u9022\u5742\u826f\u592a\uff09", "\u65e5\u8bed\u574e\u8482\u4e1d\uff08\u67da\u6728\u51c9\u9999\uff09", "\u65e5\u8bed\u91cd\u4e91\uff08\u9f50\u85e4\u58ee\u9a6c\uff09", "\u65e5\u8bed\u67ef\u83b1\uff08\u524d\u5ddd\u51c9\u5b50\uff09", "\u65e5\u8bed\u8d5b\u8bfa\uff08\u5165\u91ce\u81ea\u7531\uff09", "\u65e5\u8bed\u6234\u56e0\u65af\u96f7\u5e03\uff08\u6d25\u7530\u5065\u6b21\u90ce\uff09", "\u65e5\u8bed\u8fea\u5362\u514b\uff08\u5c0f\u91ce\u8d24\u7ae0\uff09", "\u65e5\u8bed\u8fea\u5965\u5a1c\uff08\u4e95\u6cfd\u8bd7\u7ec7\uff09", "\u65e5\u8bed\u591a\u8389\uff08\u91d1\u7530\u670b\u5b50\uff09", "\u65e5\u8bed\u4f18\u83c8\uff08\u4f50\u85e4\u5229\u5948\uff09", "\u65e5\u8bed\u83f2\u8c22\u5c14\uff08\u5185\u7530\u771f\u793c\uff09", "\u65e5\u8bed\u7518\u96e8\uff08\u4e0a\u7530\u4e3d\u5948\uff09", "\u65e5\u8bed\uff08\u7560\u4e2d\u7950\uff09", "\u65e5\u8bed\u9e7f\u91ce\u9662\u5e73\u85cf\uff08\u4e95\u53e3\u7950\u4e00\uff09", "\u65e5\u8bed\u7a7a\uff08\u5800\u6c5f\u77ac\uff09", "\u65e5\u8bed\u8367\uff08\u60a0\u6728\u78a7\uff09", "\u65e5\u8bed\u80e1\u6843\uff08\u9ad8\u6865\u674e\u4f9d\uff09", "\u65e5\u8bed\u4e00\u6597\uff08\u897f\u5ddd\u8d35\u6559\uff09", "\u65e5\u8bed\u51ef\u4e9a\uff08\u9e1f\u6d77\u6d69\u8f85\uff09", "\u65e5\u8bed\u4e07\u53f6\uff08\u5c9b\u5d0e\u4fe1\u957f\uff09", "\u65e5\u8bed\u523b\u6674\uff08\u559c\u591a\u6751\u82f1\u68a8\uff09", "\u65e5\u8bed\u53ef\u8389\uff08\u4e45\u91ce\u7f8e\u54b2\uff09", "\u65e5\u8bed\u5fc3\u6d77\uff08\u4e09\u68ee\u94c3\u5b50\uff09", "\u65e5\u8bed\u4e5d\u6761\u88df\u7f57\uff08\u6fd1\u6237\u9ebb\u6c99\u7f8e\uff09", "\u65e5\u8bed\u4e3d\u838e\uff08\u7530\u4e2d\u7406\u60e0\uff09", "\u65e5\u8bed\u83ab\u5a1c\uff08\u5c0f\u539f\u597d\u7f8e\uff09", "\u65e5\u8bed\u7eb3\u897f\u59b2\uff08\u7530\u6751\u7531\u52a0\u8389\uff09", "\u65e5\u8bed\u59ae\u9732\uff08\u91d1\u5143\u5bff\u5b50\uff09", "\u65e5\u8bed\u51dd\u5149\uff08\u5927\u539f\u6c99\u8036\u9999\uff09", "\u65e5\u8bed\u8bfa\u827e\u5c14\uff08\u9ad8\u5c3e\u594f\u97f3\uff09", "\u65e5\u8bed\u5965\u5179\uff08\u589e\u8c37\u5eb7\u7eaa\uff09", "\u65e5\u8bed\u6d3e\u8499\uff08\u53e4\u8d3a\u8475\uff09", "\u65e5\u8bed\u7434\uff08\u658b\u85e4\u5343\u548c\uff09", "\u65e5\u8bed\u4e03\u4e03\uff08\u7530\u6751\u7531\u52a0\u8389\uff09", "\u65e5\u8bed\u96f7\u7535\u5c06\u519b\uff08\u6cfd\u57ce\u7f8e\u96ea\uff09", "\u65e5\u8bed\u96f7\u6cfd\uff08\u5185\u5c71\u6602\u8f89\uff09", "\u65e5\u8bed\u7f57\u838e\u8389\u4e9a\uff08\u52a0\u9688\u4e9a\u8863\uff09", "\u65e5\u8bed\u65e9\u67da\uff08\u6d32\u5d0e\u7eeb\uff09", "\u65e5\u8bed\u6563\u5175\uff08\u67ff\u539f\u5f7b\u4e5f\uff09", "\u65e5\u8bed\u7533\u9e64\uff08\u5ddd\u6f84\u7eeb\u5b50\uff09", "\u65e5\u8bed\u4e45\u5c90\u5fcd\uff08\u6c34\u6865\u9999\u7ec7\uff09", "\u65e5\u8bed\u5973\u58eb\uff08\u5e84\u5b50\u88d5\u8863\uff09", "\u65e5\u8bed\u7802\u7cd6\uff08\u85e4\u7530\u831c\uff09", "\u65e5\u8bed\u8fbe\u8fbe\u5229\u4e9a\uff08\u6728\u6751\u826f\u5e73\uff09", "\u65e5\u8bed\u6258\u9a6c\uff08\u68ee\u7530\u6210\u4e00\uff09", "\u65e5\u8bed\u63d0\u7eb3\u91cc\uff08\u5c0f\u6797\u6c99\u82d7\uff09", "\u65e5\u8bed\u6e29\u8fea\uff08\u6751\u6fd1\u6b65\uff09", "\u65e5\u8bed\u9999\u83f1\uff08\u5c0f\u6cfd\u4e9a\u674e\uff09", "\u65e5\u8bed\u9b48\uff08\u677e\u5188\u796f\u4e1e\uff09", "\u65e5\u8bed\u884c\u79cb\uff08\u7686\u5ddd\u7eaf\u5b50\uff09", "\u65e5\u8bed\u8f9b\u7131\uff08\u9ad8\u6865\u667a\u79cb\uff09", "\u65e5\u8bed\u516b\u91cd\u795e\u5b50\uff08\u4f50\u4ed3\u7eeb\u97f3\uff09", "\u65e5\u8bed\u70df\u7eef\uff08\u82b1\u5b88\u7531\u7f8e\u91cc\uff09", "\u65e5\u8bed\u591c\u5170\uff08\u8fdc\u85e4\u7eeb\uff09", "\u65e5\u8bed\u5bb5\u5bab\uff08\u690d\u7530\u4f73\u5948\uff09", "\u65e5\u8bed\u4e91\u5807\uff08\u5c0f\u5ca9\u4e95\u5c0f\u9e1f\uff09", "\u65e5\u8bed\u949f\u79bb\uff08\u524d\u91ce\u667a\u662d\uff09", "\u6770\u514b", "\u963f\u5409", "\u6c5f\u821f", "\u9274\u79cb", "\u5609\u4e49", "\u7eaa\u82b3", "\u666f\u6f84", "\u7ecf\u7eb6", "\u666f\u660e", "\u664b\u4f18", "\u963f\u9e20", "\u9152\u5ba2", "\u4e54\u5c14", "\u4e54\u745f\u592b", "\u7ea6\u987f", "\u4e54\u4f0a\u65af", "\u5c45\u5b89", "\u541b\u541b", "\u987a\u5409", "\u7eaf\u4e5f", "\u91cd\u4f50", "\u5927\u5c9b\u7eaf\u5e73", "\u84b2\u6cfd", "\u52d8\u89e3\u7531\u5c0f\u8def\u5065\u4e09\u90ce", "\u67ab", "\u67ab\u539f\u4e49\u5e86", "\u836b\u5c71", "\u7532\u6590\u7530\u9f8d\u99ac", "\u6d77\u6597", "\u60df\u795e\u6674\u4e4b\u4ecb", "\u9e7f\u91ce\u5948\u5948", "\u5361\u7435\u8389\u4e9a", "\u51ef\u745f\u7433", "\u52a0\u85e4\u4fe1\u609f", "\u52a0\u85e4\u6d0b\u5e73", "\u80dc\u5bb6", "\u8305\u847a\u4e00\u5e86", "\u548c\u662d", "\u4e00\u6b63", "\u4e00\u9053", "\u6842\u4e00", "\u5e86\u6b21\u90ce", "\u963f\u8d24", "\u5065\u53f8", "\u5065\u6b21\u90ce", "\u5065\u4e09\u90ce", "\u5929\u7406", "\u6740\u624ba", "\u6740\u624bb", "\u6728\u5357\u674f\u5948", "\u6728\u6751", "\u56fd\u738b", "\u6728\u4e0b", "\u5317\u6751", "\u6e05\u60e0", "\u6e05\u4eba", "\u514b\u5217\u95e8\u7279", "\u9a91\u58eb", "\u5c0f\u6797", "\u5c0f\u6625", "\u5eb7\u62c9\u5fb7", "\u5927\u8089\u4e38", "\u7434\u7f8e", "\u5b8f\u4e00", "\u5eb7\u4ecb", "\u5e78\u5fb7", "\u9ad8\u5584", "\u68a2", "\u514b\u7f57\u7d22", "\u4e45\u4fdd", "\u4e5d\u6761\u9570\u6cbb", "\u4e45\u6728\u7530", "\u6606\u94a7", "\u83ca\u5730\u541b", "\u4e45\u5229\u987b", "\u9ed1\u7530", "\u9ed1\u6cfd\u4eac\u4e4b\u4ecb", "\u54cd\u592a", "\u5c9a\u59d0", "\u5170\u6eaa", "\u6f9c\u9633", "\u52b3\u4f26\u65af", "\u4e50\u660e", "\u83b1\u8bfa", "\u83b2", "\u826f\u5b50", "\u674e\u5f53", "\u674e\u4e01", "\u5c0f\u4e50", "\u7075", "\u5c0f\u73b2", "\u7433\u7405a", "\u7433\u7405b", "\u5c0f\u5f6c", "\u5c0f\u5fb7", "\u5c0f\u697d", "\u5c0f\u9f99", "\u5c0f\u5434", "\u5c0f\u5434\u7684\u8bb0\u5fc6", "\u7406\u6b63", "\u963f\u9f99", "\u5362\u5361", "\u6d1b\u6210", "\u7f57\u5de7", "\u5317\u98ce\u72fc", "\u5362\u6b63", "\u840d\u59e5\u59e5", "\u524d\u7530", "\u771f\u663c", "\u9ebb\u7eaa", "\u771f", "\u611a\u4eba\u4f17-\u9a6c\u514b\u897f\u59c6", "\u5973\u6027a", "\u5973\u6027b", "\u5973\u6027a\u7684\u8ddf\u968f\u8005", "\u963f\u5b88", "\u739b\u683c\u4e3d\u7279", "\u771f\u7406", "\u739b\u4e54\u4e3d", "\u739b\u6587", "\u6b63\u80dc", "\u660c\u4fe1", "\u5c06\u53f8", "\u6b63\u4eba", "\u8def\u7237", "\u8001\u7ae0", "\u677e\u7530", "\u677e\u672c", "\u677e\u6d66", "\u677e\u5742", "\u8001\u5b5f", "\u5b5f\u4e39", "\u5546\u4eba\u968f\u4ece", "\u4f20\u4ee4\u5175", "\u7c73\u6b47\u5c14", "\u5fa1\u8206\u6e90\u4e00\u90ce", "\u5fa1\u8206\u6e90\u6b21\u90ce", "\u5343\u5ca9\u519b\u6559\u5934", "\u5343\u5ca9\u519b\u58eb\u5175", "\u660e\u535a", "\u660e\u4fca", "\u7f8e\u94c3", "\u7f8e\u548c", "\u963f\u5e78", "\u524a\u6708\u7b51\u9633\u771f\u541b", "\u94b1\u773c\u513f", "\u68ee\u5f66", "\u5143\u52a9", "\u7406\u6c34\u53e0\u5c71\u771f\u541b", "\u7406\u6c34\u758a\u5c71\u771f\u541b", "\u6731\u8001\u677f", "\u6728\u6728", "\u6751\u4e0a", "\u6751\u7530", "\u6c38\u91ce", "\u957f\u91ce\u539f\u9f99\u4e4b\u4ecb", "\u957f\u6fd1", "\u4e2d\u91ce\u5fd7\u4e43", "\u83dc\u83dc\u5b50", "\u6960\u6960", "\u6210\u6fd1", "\u963f\u5185", "\u5b81\u7984", "\u725b\u5fd7", "\u4fe1\u535a", "\u4f38\u592b", "\u91ce\u65b9", "\u8bfa\u62c9", "\u7eaa\u9999", "\u8bfa\u66fc", "\u4fee\u5973", "\u7eaf\u6c34\u7cbe\u7075", "\u5c0f\u5ddd", "\u5c0f\u4ed3\u6faa", "\u5188\u6797", "\u5188\u5d0e\u7ed8\u91cc\u9999", "\u5188\u5d0e\u9646\u6597", "\u5965\u62c9\u592b", "\u8001\u79d1", "\u9b3c\u5a46\u5a46", "\u5c0f\u91ce\u5bfa", "\u5927\u6cb3\u539f\u4e94\u53f3\u536b\u95e8", "\u5927\u4e45\u4fdd\u5927\u4ecb", "\u5927\u68ee", "\u5927\u52a9", "\u5965\u7279", "\u6d3e\u8499", "\u6d3e\u84992", "\u75c5\u4ebaa", "\u75c5\u4ebab", "\u5df4\u987f", "\u6d3e\u6069", "\u670b\u4e49", "\u56f4\u89c2\u7fa4\u4f17", "\u56f4\u89c2\u7fa4\u4f17a", "\u56f4\u89c2\u7fa4\u4f17b", "\u56f4\u89c2\u7fa4\u4f17c", "\u56f4\u89c2\u7fa4\u4f17d", "\u56f4\u89c2\u7fa4\u4f17e", "\u94dc\u96c0", "\u963f\u80a5", "\u5174\u53d4", "\u8001\u5468\u53d4", "\u516c\u4e3b", "\u5f7c\u5f97", "\u4e7e\u5b50", "\u828a\u828a", "\u4e7e\u73ae", "\u7eee\u547d", "\u675e\u5e73", "\u79cb\u6708", "\u6606\u6069", "\u96f7\u7535\u5f71", "\u5170\u9053\u5c14", "\u96f7\u8499\u5fb7", "\u5192\u5931\u7684\u5e15\u62c9\u5fb7", "\u4f36\u4e00", "\u73b2\u82b1", "\u963f\u4ec1", "\u5bb6\u81e3\u4eec", "\u68a8\u7ed8", "\u8363\u6c5f", "\u620e\u4e16", "\u6d6a\u4eba", "\u7f57\u4f0a\u65af", "\u5982\u610f", "\u51c9\u5b50", "\u5f69\u9999", "\u9152\u4e95", "\u5742\u672c", "\u6714\u6b21\u90ce", "\u6b66\u58eba", "\u6b66\u58ebb", "\u6b66\u58ebc", "\u6b66\u58ebd", "\u73ca\u745a", "\u4e09\u7530", "\u838e\u62c9", "\u7b39\u91ce", "\u806a\u7f8e", "\u806a", "\u5c0f\u767e\u5408", "\u6563\u5175", "\u5bb3\u6015\u7684\u5c0f\u5218", "\u8212\u4f2f\u7279", "\u8212\u8328", "\u6d77\u9f99", "\u4e16\u5b50", "\u8c22\u5c14\u76d6", "\u5bb6\u4e01", "\u5546\u534e", "\u6c99\u5bc5", "\u963f\u5347", "\u67f4\u7530", "\u963f\u8302", "\u5f0f\u5927\u5c06", "\u6e05\u6c34", "\u5fd7\u6751\u52d8\u5175\u536b", "\u65b0\u4e4b\u4e1e", "\u5fd7\u7ec7", "\u77f3\u5934", "\u8bd7\u7fbd", "\u8bd7\u7b60", "\u77f3\u58ee", "\u7fd4\u592a", "\u6b63\u4e8c", "\u5468\u5e73", "\u8212\u6768", "\u9f50\u683c\u8299\u4e3d\u96c5", "\u5973\u58eb", "\u601d\u52e4", "\u516d\u6307\u4e54\u745f", "\u611a\u4eba\u4f17\u5c0f\u5175d", "\u611a\u4eba\u4f17\u5c0f\u5175a", "\u611a\u4eba\u4f17\u5c0f\u5175b", "\u611a\u4eba\u4f17\u5c0f\u5175c", "\u5434\u8001\u4e94", "\u5434\u8001\u4e8c", "\u6ed1\u5934\u9b3c", "\u8a00\u7b11", "\u5434\u8001\u4e03", "\u58eb\u5175h", "\u58eb\u5175i", "\u58eb\u5175a", "\u58eb\u5175b", "\u58eb\u5175c", "\u58eb\u5175d", "\u58eb\u5175e", "\u58eb\u5175f", "\u58eb\u5175g", "\u594f\u592a", "\u65af\u5766\u5229", "\u6387\u661f\u652b\u8fb0\u5929\u541b", "\u5c0f\u5934", "\u5927\u6b66", "\u9676\u4e49\u9686", "\u6749\u672c", "\u82cf\u897f", "\u5acc\u7591\u4ebaa", "\u5acc\u7591\u4ebab", "\u5acc\u7591\u4ebac", "\u5acc\u7591\u4ebad", "\u65af\u4e07", "\u5251\u5ba2a", "\u5251\u5ba2b", "\u963f\u4e8c", "\u5fe0\u80dc", "\u5fe0\u592b", "\u963f\u656c", "\u5b5d\u5229", "\u9e70\u53f8\u8fdb", "\u9ad8\u5c71", "\u4e5d\u6761\u5b5d\u884c", "\u6bc5", "\u7af9\u5185", "\u62d3\u771f", "\u5353\u4e5f", "\u592a\u90ce\u4e38", "\u6cf0\u52d2", "\u624b\u5c9b", "\u54f2\u5e73", "\u54f2\u592b", "\u6258\u514b", "\u5927boss", "\u963f\u5f3a", "\u6258\u5c14\u5fb7\u62c9", "\u65c1\u89c2\u8005", "\u5929\u6210", "\u963f\u5927", "\u8482\u739b\u4e4c\u65af", "\u63d0\u7c73", "\u6237\u7530", "\u963f\u4e09", "\u4e00\u8d77\u7684\u4eba", "\u5fb7\u7530", "\u5fb7\u957f", "\u667a\u6811", "\u5229\u5f66", "\u80d6\u4e4e\u4e4e\u7684\u65c5\u884c\u8005", "\u85cf\u5b9d\u4ebaa", "\u85cf\u5b9d\u4ebab", "\u85cf\u5b9d\u4ebac", "\u85cf\u5b9d\u4ebad", "\u963f\u7947", "\u6052\u96c4", "\u9732\u5b50", "\u8bdd\u5267\u56e2\u56e2\u957f", "\u5185\u6751", "\u4e0a\u91ce", "\u4e0a\u6749", "\u8001\u6234", "\u8001\u9ad8", "\u8001\u8d3e", "\u8001\u58a8", "\u8001\u5b59", "\u5929\u67a2\u661f", "\u8001\u4e91", "\u6709\u4e50\u658b", "\u4e11\u96c4", "\u4e4c\u7ef4", "\u74e6\u4eac", "\u83f2\u5c14\u6208\u9edb\u7279", "\u7ef4\u591a\u5229\u4e9a", "\u8587\u5c14", "\u74e6\u683c\u7eb3", "\u963f\u5916", "\u4f8d\u5973", "\u74e6\u62c9", "\u671b\u96c5", "\u5b9b\u70df", "\u742c\u7389", "\u6218\u58eba", "\u6218\u58ebb", "\u6e21\u8fba", "\u6e21\u90e8", "\u963f\u4f1f", "\u6587\u749f", "\u6587\u6e0a", "\u97e6\u5c14\u7eb3", "\u738b\u6273\u624b", "\u6b66\u6c9b", "\u6653\u98de", "\u8f9b\u7a0b", "\u661f\u706b", "\u661f\u7a00", "\u8f9b\u79c0", "\u79c0\u534e", "\u963f\u65ed", "\u5f90\u5218\u5e08", "\u77e2\u90e8", "\u516b\u6728", "\u5c71\u4e0a", "\u963f\u9633", "\u989c\u7b11", "\u5eb7\u660e", "\u6cf0\u4e45", "\u5b89\u6b66", "\u77e2\u7530\u5e78\u559c", "\u77e2\u7530\u8f9b\u559c", "\u4e49\u575a", "\u83ba\u513f", "\u76c8\u4e30", "\u5b9c\u5e74", "\u94f6\u674f", "\u9038\u8f69", "\u6a2a\u5c71", "\u6c38\u8d35", "\u6c38\u4e1a", "\u5609\u4e45", "\u5409\u5ddd", "\u4e49\u9ad8", "\u7528\u9ad8", "\u9633\u592a", "\u5143\u84c9", "\u73a5\u8f89", "\u6bd3\u534e", "\u6709\u9999", "\u5e78\u4e5f", "\u7531\u771f", "\u7ed3\u83dc", "\u97f5\u5b81", "\u767e\u5408", "\u767e\u5408\u534e", "\u5c24\u82cf\u6ce2\u592b", "\u88d5\u5b50", "\u60a0\u7b56", "\u60a0\u4e5f", "\u4e8e\u5ae3", "\u67da\u5b50", "\u8001\u90d1", "\u6b63\u8302", "\u5fd7\u6210", "\u82b7\u5de7", "\u77e5\u6613", "\u652f\u652f", "\u5468\u826f", "\u73e0\u51fd", "\u795d\u660e", "\u795d\u6d9b"],
+  "symbols": ["_", ",", ".", "!", "?", "-", "~", "\u2026", "A", "E", "I", "N", "O", "Q", "U", "a", "b", "d", "e", "f", "g", "h", "i", "j", "k", "l", "m", "n", "o", "p", "r", "s", "t", "u", "v", "w", "y", "z", "\u0283", "\u02a7", "\u02a6", "\u026f", "\u0279", "\u0259", "\u0265", "\u207c", "\u02b0", "`", "\u2192", "\u2193", "\u2191", " "]
+}

configs/modified_finetune_speaker.json

@@ -0,0 +1,172 @@
+{
+  "train": {
+    "log_interval": 10,
+    "eval_interval": 100,
+    "seed": 1234,
+    "epochs": 10000,
+    "learning_rate": 0.0002,
+    "betas": [
+      0.8,
+      0.99
+    ],
+    "eps": 1e-09,
+    "batch_size": 16,
+    "fp16_run": true,
+    "lr_decay": 0.999875,
+    "segment_size": 8192,
+    "init_lr_ratio": 1,
+    "warmup_epochs": 0,
+    "c_mel": 45,
+    "c_kl": 1.0
+  },
+  "data": {
+    "training_files": "final_annotation_train.txt",
+    "validation_files": "final_annotation_val.txt",
+    "text_cleaners": [
+      "chinese_cleaners"
+    ],
+    "max_wav_value": 32768.0,
+    "sampling_rate": 22050,
+    "filter_length": 1024,
+    "hop_length": 256,
+    "win_length": 1024,
+    "n_mel_channels": 80,
+    "mel_fmin": 0.0,
+    "mel_fmax": null,
+    "add_blank": true,
+    "n_speakers": 2,
+    "cleaned_text": true
+  },
+  "model": {
+    "inter_channels": 192,
+    "hidden_channels": 192,
+    "filter_channels": 768,
+    "n_heads": 2,
+    "n_layers": 6,
+    "kernel_size": 3,
+    "p_dropout": 0.1,
+    "resblock": "1",
+    "resblock_kernel_sizes": [
+      3,
+      7,
+      11
+    ],
+    "resblock_dilation_sizes": [
+      [
+        1,
+        3,
+        5
+      ],
+      [
+        1,
+        3,
+        5
+      ],
+      [
+        1,
+        3,
+        5
+      ]
+    ],
+    "upsample_rates": [
+      8,
+      8,
+      2,
+      2
+    ],
+    "upsample_initial_channel": 512,
+    "upsample_kernel_sizes": [
+      16,
+      16,
+      4,
+      4
+    ],
+    "n_layers_q": 3,
+    "use_spectral_norm": false,
+    "gin_channels": 256
+  },
+  "symbols": [
+    "_",
+    "\uff1b",
+    "\uff1a",
+    "\uff0c",
+    "\u3002",
+    "\uff01",
+    "\uff1f",
+    "-",
+    "\u201c",
+    "\u201d",
+    "\u300a",
+    "\u300b",
+    "\u3001",
+    "\uff08",
+    "\uff09",
+    "\u2026",
+    "\u2014",
+    " ",
+    "A",
+    "B",
+    "C",
+    "D",
+    "E",
+    "F",
+    "G",
+    "H",
+    "I",
+    "J",
+    "K",
+    "L",
+    "M",
+    "N",
+    "O",
+    "P",
+    "Q",
+    "R",
+    "S",
+    "T",
+    "U",
+    "V",
+    "W",
+    "X",
+    "Y",
+    "Z",
+    "a",
+    "b",
+    "c",
+    "d",
+    "e",
+    "f",
+    "g",
+    "h",
+    "i",
+    "j",
+    "k",
+    "l",
+    "m",
+    "n",
+    "o",
+    "p",
+    "q",
+    "r",
+    "s",
+    "t",
+    "u",
+    "v",
+    "w",
+    "x",
+    "y",
+    "z",
+    "1",
+    "2",
+    "3",
+    "4",
+    "5",
+    "0",
+    "\uff22",
+    "\uff30"
+  ],
+  "speakers": {
+    "dingzhen": 0,
+    "taffy": 1
+  }
+}

custom_character_voice/22050.txt

@@ -0,0 +1 @@
+

data_utils.py

@@ -0,0 +1,529 @@
+import time
+import os
+import random
+import numpy as np
+import torch
+import torch.utils.data
+
+import commons
+from mel_processing import spectrogram_torch, mel_spectrogram_torch, spec_to_mel_torch
+from utils import load_wav_to_torch, load_filepaths_and_text
+from text import text_to_sequence, cleaned_text_to_sequence
+
+
+class TextAudioLoader(torch.utils.data.Dataset):
+    """
+    1) loads audio, text pairs
+    2) normalizes text and converts them to sequences of integers
+    3) computes spectrograms from audio files.
+    """
+
+    def __init__(self, audiopaths_and_text, hparams):
+        self.hparams = hparams
+        self.audiopaths_and_text = load_filepaths_and_text(audiopaths_and_text)
+        self.text_cleaners = hparams.text_cleaners
+        self.max_wav_value = hparams.max_wav_value
+        self.sampling_rate = hparams.sampling_rate
+        self.filter_length = hparams.filter_length
+        self.hop_length = hparams.hop_length
+        self.win_length = hparams.win_length
+        self.sampling_rate = hparams.sampling_rate
+
+        self.use_mel_spec_posterior = getattr(
+            hparams, "use_mel_posterior_encoder", False
+        )
+        if self.use_mel_spec_posterior:
+            self.n_mel_channels = getattr(hparams, "n_mel_channels", 80)
+        self.cleaned_text = getattr(hparams, "cleaned_text", False)
+
+        self.add_blank = hparams.add_blank
+        self.min_text_len = getattr(hparams, "min_text_len", 1)
+        self.max_text_len = getattr(hparams, "max_text_len", 190)
+
+        random.seed(1234)
+        random.shuffle(self.audiopaths_and_text)
+        self._filter()
+
+    def _filter(self):
+        """
+        Filter text & store spec lengths
+        """
+        # Store spectrogram lengths for Bucketing
+        # wav_length ~= file_size / (wav_channels * Bytes per dim) = file_size / (1 * 2)
+        # spec_length = wav_length // hop_length
+
+        audiopaths_and_text_new = []
+        lengths = []
+        for audiopath, text in self.audiopaths_and_text:
+            if self.min_text_len <= len(text) and len(text) <= self.max_text_len:
+                audiopaths_and_text_new.append([audiopath, text])
+                lengths.append(os.path.getsize(audiopath) // (2 * self.hop_length))
+        self.audiopaths_and_text = audiopaths_and_text_new
+        self.lengths = lengths
+
+    def get_audio_text_pair(self, audiopath_and_text):
+        # separate filename and text
+        audiopath, text = audiopath_and_text[0], audiopath_and_text[1]
+        text = self.get_text(text)
+        spec, wav = self.get_audio(audiopath)
+        return (text, spec, wav)
+
+    def get_audio(self, filename):
+        # TODO : if linear spec exists convert to mel from existing linear spec
+        audio, sampling_rate = load_wav_to_torch(filename)
+        if sampling_rate != self.sampling_rate:
+            raise ValueError(
+                "{} {} SR doesn't match target {} SR".format(
+                    sampling_rate, self.sampling_rate
+                )
+            )
+        audio_norm = audio / self.max_wav_value
+        audio_norm = audio_norm.unsqueeze(0)
+        spec_filename = filename.replace(".wav", ".spec.pt")
+        if self.use_mel_spec_posterior:
+            spec_filename = spec_filename.replace(".spec.pt", ".mel.pt")
+        if os.path.exists(spec_filename):
+            spec = torch.load(spec_filename)
+        else:
+            if self.use_mel_spec_posterior:
+                """TODO : (need verification)
+                if linear spec exists convert to
+                mel from existing linear spec (uncomment below lines)"""
+                # if os.path.exists(filename.replace(".wav", ".spec.pt")):
+                #     # spec, n_fft, num_mels, sampling_rate, fmin, fmax
+                #     spec = spec_to_mel_torch(
+                #         torch.load(filename.replace(".wav", ".spec.pt")),
+                #         self.filter_length, self.n_mel_channels, self.sampling_rate,
+                #         self.hparams.mel_fmin, self.hparams.mel_fmax)
+                spec = mel_spectrogram_torch(
+                    audio_norm,
+                    self.filter_length,
+                    self.n_mel_channels,
+                    self.sampling_rate,
+                    self.hop_length,
+                    self.win_length,
+                    self.hparams.mel_fmin,
+                    self.hparams.mel_fmax,
+                    center=False,
+                )
+            else:
+                spec = spectrogram_torch(
+                    audio_norm,
+                    self.filter_length,
+                    self.sampling_rate,
+                    self.hop_length,
+                    self.win_length,
+                    center=False,
+                )
+            spec = torch.squeeze(spec, 0)
+            torch.save(spec, spec_filename)
+        return spec, audio_norm
+
+    def get_text(self, text):
+        if self.cleaned_text:
+            text_norm = cleaned_text_to_sequence(text)
+        else:
+            text_norm = text_to_sequence(text, self.text_cleaners)
+        if self.add_blank:
+            text_norm = commons.intersperse(text_norm, 0)
+        text_norm = torch.LongTensor(text_norm)
+        return text_norm
+
+    def __getitem__(self, index):
+        return self.get_audio_text_pair(self.audiopaths_and_text[index])
+
+    def __len__(self):
+        return len(self.audiopaths_and_text)
+
+
+class TextAudioCollate:
+    """Zero-pads model inputs and targets"""
+
+    def __init__(self, return_ids=False):
+        self.return_ids = return_ids
+
+    def __call__(self, batch):
+        """Collate's training batch from normalized text and aduio
+        PARAMS
+        ------
+        batch: [text_normalized, spec_normalized, wav_normalized]
+        """
+        # Right zero-pad all one-hot text sequences to max input length
+        _, ids_sorted_decreasing = torch.sort(
+            torch.LongTensor([x[1].size(1) for x in batch]), dim=0, descending=True
+        )
+
+        max_text_len = max([len(x[0]) for x in batch])
+        max_spec_len = max([x[1].size(1) for x in batch])
+        max_wav_len = max([x[2].size(1) for x in batch])
+
+        text_lengths = torch.LongTensor(len(batch))
+        spec_lengths = torch.LongTensor(len(batch))
+        wav_lengths = torch.LongTensor(len(batch))
+
+        text_padded = torch.LongTensor(len(batch), max_text_len)
+        spec_padded = torch.FloatTensor(len(batch), batch[0][1].size(0), max_spec_len)
+        wav_padded = torch.FloatTensor(len(batch), 1, max_wav_len)
+        text_padded.zero_()
+        spec_padded.zero_()
+        wav_padded.zero_()
+        for i in range(len(ids_sorted_decreasing)):
+            row = batch[ids_sorted_decreasing[i]]
+
+            text = row[0]
+            text_padded[i, : text.size(0)] = text
+            text_lengths[i] = text.size(0)
+
+            spec = row[1]
+            spec_padded[i, :, : spec.size(1)] = spec
+            spec_lengths[i] = spec.size(1)
+
+            wav = row[2]
+            wav_padded[i, :, : wav.size(1)] = wav
+            wav_lengths[i] = wav.size(1)
+
+        if self.return_ids:
+            return (
+                text_padded,
+                text_lengths,
+                spec_padded,
+                spec_lengths,
+                wav_padded,
+                wav_lengths,
+                ids_sorted_decreasing,
+            )
+        return (
+            text_padded,
+            text_lengths,
+            spec_padded,
+            spec_lengths,
+            wav_padded,
+            wav_lengths,
+        )
+
+
+"""Multi speaker version"""
+
+
+class TextAudioSpeakerLoader(torch.utils.data.Dataset):
+    """
+    1) loads audio, speaker_id, text pairs
+    2) normalizes text and converts them to sequences of integers
+    3) computes spectrograms from audio files.
+    """
+
+    def __init__(self, audiopaths_sid_text, hparams):
+        self.hparams = hparams
+        self.audiopaths_sid_text = load_filepaths_and_text(audiopaths_sid_text)
+        self.text_cleaners = hparams.text_cleaners
+        self.max_wav_value = hparams.max_wav_value
+        self.sampling_rate = hparams.sampling_rate
+        self.filter_length = hparams.filter_length
+        self.hop_length = hparams.hop_length
+        self.win_length = hparams.win_length
+        self.sampling_rate = hparams.sampling_rate
+
+        self.use_mel_spec_posterior = getattr(
+            hparams, "use_mel_posterior_encoder", False
+        )
+        if self.use_mel_spec_posterior:
+            self.n_mel_channels = getattr(hparams, "n_mel_channels", 80)
+        self.cleaned_text = getattr(hparams, "cleaned_text", False)
+
+        self.add_blank = hparams.add_blank
+        self.min_text_len = getattr(hparams, "min_text_len", 1)
+        self.max_text_len = getattr(hparams, "max_text_len", 190)
+
+        random.seed(1234)
+        random.shuffle(self.audiopaths_sid_text)
+        self._filter()
+
+    def _filter(self):
+        """
+        Filter text & store spec lengths
+        """
+        # Store spectrogram lengths for Bucketing
+        # wav_length ~= file_size / (wav_channels * Bytes per dim) = file_size / (1 * 2)
+        # spec_length = wav_length // hop_length
+
+        audiopaths_sid_text_new = []
+        lengths = []
+        for audiopath, sid, text in self.audiopaths_sid_text:
+            if self.min_text_len <= len(text) and len(text) <= self.max_text_len:
+                audiopaths_sid_text_new.append([audiopath, sid, text])
+                lengths.append(os.path.getsize(audiopath) // (2 * self.hop_length))
+        self.audiopaths_sid_text = audiopaths_sid_text_new
+        self.lengths = lengths
+
+    def get_audio_text_speaker_pair(self, audiopath_sid_text):
+        # separate filename, speaker_id and text
+        audiopath, sid, text = (
+            audiopath_sid_text[0],
+            audiopath_sid_text[1],
+            audiopath_sid_text[2],
+        )
+        text = self.get_text(text)
+        spec, wav = self.get_audio(audiopath)
+        sid = self.get_sid(sid)
+        return (text, spec, wav, sid)
+
+    def get_audio(self, filename):
+        # TODO : if linear spec exists convert to mel from existing linear spec
+        audio, sampling_rate = load_wav_to_torch(filename)
+        if sampling_rate != self.sampling_rate:
+            raise ValueError(
+                "{} {} SR doesn't match target {} SR".format(
+                    sampling_rate, self.sampling_rate
+                )
+            )
+        audio_norm = audio / self.max_wav_value
+        audio_norm = audio_norm.unsqueeze(0)
+        spec_filename = filename.replace(".wav", ".spec.pt")
+        if self.use_mel_spec_posterior:
+            spec_filename = spec_filename.replace(".spec.pt", ".mel.pt")
+        if os.path.exists(spec_filename):
+            spec = torch.load(spec_filename)
+        else:
+            if self.use_mel_spec_posterior:
+                """TODO : (need verification)
+                if linear spec exists convert to
+                mel from existing linear spec (uncomment below lines)"""
+                # if os.path.exists(filename.replace(".wav", ".spec.pt")):
+                #     # spec, n_fft, num_mels, sampling_rate, fmin, fmax
+                #     spec = spec_to_mel_torch(
+                #         torch.load(filename.replace(".wav", ".spec.pt")),
+                #         self.filter_length, self.n_mel_channels, self.sampling_rate,
+                #         self.hparams.mel_fmin, self.hparams.mel_fmax)
+                spec = mel_spectrogram_torch(
+                    audio_norm,
+                    self.filter_length,
+                    self.n_mel_channels,
+                    self.sampling_rate,
+                    self.hop_length,
+                    self.win_length,
+                    self.hparams.mel_fmin,
+                    self.hparams.mel_fmax,
+                    center=False,
+                )
+            else:
+                spec = spectrogram_torch(
+                    audio_norm,
+                    self.filter_length,
+                    self.sampling_rate,
+                    self.hop_length,
+                    self.win_length,
+                    center=False,
+                )
+            spec = torch.squeeze(spec, 0)
+            torch.save(spec, spec_filename)
+        return spec, audio_norm
+
+    def get_text(self, text):
+        if self.cleaned_text:
+            text_norm = cleaned_text_to_sequence(text)
+        else:
+            text_norm = text_to_sequence(text, self.text_cleaners)
+        if self.add_blank:
+            text_norm = commons.intersperse(text_norm, 0)
+        text_norm = torch.LongTensor(text_norm)
+        return text_norm
+
+    def get_sid(self, sid):
+        sid = torch.LongTensor([int(sid)])
+        return sid
+
+    def __getitem__(self, index):
+        return self.get_audio_text_speaker_pair(self.audiopaths_sid_text[index])
+
+    def __len__(self):
+        return len(self.audiopaths_sid_text)
+
+
+class TextAudioSpeakerCollate:
+    """Zero-pads model inputs and targets"""
+
+    def __init__(self, return_ids=False):
+        self.return_ids = return_ids
+
+    def __call__(self, batch):
+        """Collate's training batch from normalized text, audio and speaker identities
+        PARAMS
+        ------
+        batch: [text_normalized, spec_normalized, wav_normalized, sid]
+        """
+        # Right zero-pad all one-hot text sequences to max input length
+        _, ids_sorted_decreasing = torch.sort(
+            torch.LongTensor([x[1].size(1) for x in batch]), dim=0, descending=True
+        )
+
+        max_text_len = max([len(x[0]) for x in batch])
+        max_spec_len = max([x[1].size(1) for x in batch])
+        max_wav_len = max([x[2].size(1) for x in batch])
+
+        text_lengths = torch.LongTensor(len(batch))
+        spec_lengths = torch.LongTensor(len(batch))
+        wav_lengths = torch.LongTensor(len(batch))
+        sid = torch.LongTensor(len(batch))
+
+        text_padded = torch.LongTensor(len(batch), max_text_len)
+        spec_padded = torch.FloatTensor(len(batch), batch[0][1].size(0), max_spec_len)
+        wav_padded = torch.FloatTensor(len(batch), 1, max_wav_len)
+        text_padded.zero_()
+        spec_padded.zero_()
+        wav_padded.zero_()
+        for i in range(len(ids_sorted_decreasing)):
+            row = batch[ids_sorted_decreasing[i]]
+
+            text = row[0]
+            text_padded[i, : text.size(0)] = text
+            text_lengths[i] = text.size(0)
+
+            spec = row[1]
+            spec_padded[i, :, : spec.size(1)] = spec
+            spec_lengths[i] = spec.size(1)
+
+            wav = row[2]
+            wav_padded[i, :, : wav.size(1)] = wav
+            wav_lengths[i] = wav.size(1)
+
+            sid[i] = row[3]
+
+        if self.return_ids:
+            return (
+                text_padded,
+                text_lengths,
+                spec_padded,
+                spec_lengths,
+                wav_padded,
+                wav_lengths,
+                sid,
+                ids_sorted_decreasing,
+            )
+        return (
+            text_padded,
+            text_lengths,
+            spec_padded,
+            spec_lengths,
+            wav_padded,
+            wav_lengths,
+            sid,
+        )
+
+
+class DistributedBucketSampler(torch.utils.data.distributed.DistributedSampler):
+    """
+    Maintain similar input lengths in a batch.
+    Length groups are specified by boundaries.
+    Ex) boundaries = [b1, b2, b3] -> any batch is included either {x | b1 < length(x) <=b2} or {x | b2 < length(x) <= b3}.
+
+    It removes samples which are not included in the boundaries.
+    Ex) boundaries = [b1, b2, b3] -> any x s.t. length(x) <= b1 or length(x) > b3 are discarded.
+    """
+
+    def __init__(
+        self,
+        dataset,
+        batch_size,
+        boundaries,
+        num_replicas=None,
+        rank=None,
+        shuffle=True,
+    ):
+        super().__init__(dataset, num_replicas=num_replicas, rank=rank, shuffle=shuffle)
+        self.lengths = dataset.lengths
+        self.batch_size = batch_size
+        self.boundaries = boundaries
+
+        self.buckets, self.num_samples_per_bucket = self._create_buckets()
+        self.total_size = sum(self.num_samples_per_bucket)
+        self.num_samples = self.total_size // self.num_replicas
+
+    def _create_buckets(self):
+        buckets = [[] for _ in range(len(self.boundaries) - 1)]
+        for i in range(len(self.lengths)):
+            length = self.lengths[i]
+            idx_bucket = self._bisect(length)
+            if idx_bucket != -1:
+                buckets[idx_bucket].append(i)
+
+        for i in range(len(buckets) - 1, 0, -1):
+            if len(buckets[i]) == 0:
+                buckets.pop(i)
+                self.boundaries.pop(i + 1)
+
+        num_samples_per_bucket = []
+        for i in range(len(buckets)):
+            len_bucket = len(buckets[i])
+            total_batch_size = self.num_replicas * self.batch_size
+            rem = (
+                total_batch_size - (len_bucket % total_batch_size)
+            ) % total_batch_size
+            num_samples_per_bucket.append(len_bucket + rem)
+        return buckets, num_samples_per_bucket
+
+    def __iter__(self):
+        # deterministically shuffle based on epoch
+        g = torch.Generator()
+        g.manual_seed(self.epoch)
+
+        indices = []
+        if self.shuffle:
+            for bucket in self.buckets:
+                indices.append(torch.randperm(len(bucket), generator=g).tolist())
+        else:
+            for bucket in self.buckets:
+                indices.append(list(range(len(bucket))))
+
+        batches = []
+        for i in range(len(self.buckets)):
+            bucket = self.buckets[i]
+            len_bucket = len(bucket)
+            ids_bucket = indices[i]
+            num_samples_bucket = self.num_samples_per_bucket[i]
+
+            # add extra samples to make it evenly divisible
+            rem = num_samples_bucket - len_bucket
+            ids_bucket = (
+                ids_bucket
+                + ids_bucket * (rem // len_bucket)
+                + ids_bucket[: (rem % len_bucket)]
+            )
+
+            # subsample
+            ids_bucket = ids_bucket[self.rank :: self.num_replicas]
+
+            # batching
+            for j in range(len(ids_bucket) // self.batch_size):
+                batch = [
+                    bucket[idx]
+                    for idx in ids_bucket[
+                        j * self.batch_size : (j + 1) * self.batch_size
+                    ]
+                ]
+                batches.append(batch)
+
+        if self.shuffle:
+            batch_ids = torch.randperm(len(batches), generator=g).tolist()
+            batches = [batches[i] for i in batch_ids]
+        self.batches = batches
+
+        assert len(self.batches) * self.batch_size == self.num_samples
+        return iter(self.batches)
+
+    def _bisect(self, x, lo=0, hi=None):
+        if hi is None:
+            hi = len(self.boundaries) - 1
+
+        if hi > lo:
+            mid = (hi + lo) // 2
+            if self.boundaries[mid] < x and x <= self.boundaries[mid + 1]:
+                return mid
+            elif x <= self.boundaries[mid]:
+                return self._bisect(x, lo, mid)
+            else:
+                return self._bisect(x, mid + 1, hi)
+        else:
+            return -1
+
+    def __len__(self):
+        return self.num_samples // self.batch_size

export_onnx.py

@@ -0,0 +1,110 @@
+import argparse
+from pathlib import Path
+from typing import Optional
+
+import torch
+
+import utils
+from models import SynthesizerTrn
+from text.symbols import symbols
+
+OPSET_VERSION = 15
+
+
+def main() -> None:
+    torch.manual_seed(1234)
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--model-path", required=True, help="Path to model weights (.pth)"
+    )
+    parser.add_argument(
+        "--config-path", required=True, help="Path to model config (.json)"
+    )
+    parser.add_argument("--output", required=True, help="Path to output model (.onnx)")
+
+    args = parser.parse_args()
+
+    args.model_path = Path(args.model_path)
+    args.config_path = Path(args.config_path)
+    args.output = Path(args.output)
+    args.output.parent.mkdir(parents=True, exist_ok=True)
+
+    hps = utils.get_hparams_from_file(args.config_path)
+
+    if (
+        "use_mel_posterior_encoder" in hps.model.keys()
+        and hps.model.use_mel_posterior_encoder == True
+    ):
+        print("Using mel posterior encoder for VITS2")
+        posterior_channels = 80  # vits2
+        hps.data.use_mel_posterior_encoder = True
+    else:
+        print("Using lin posterior encoder for VITS1")
+        posterior_channels = hps.data.filter_length // 2 + 1
+        hps.data.use_mel_posterior_encoder = False
+
+    model_g = SynthesizerTrn(
+        len(symbols),
+        posterior_channels,
+        hps.train.segment_size // hps.data.hop_length,
+        n_speakers=hps.data.n_speakers,
+        **hps.model,
+    )
+
+    _ = model_g.eval()
+
+    _ = utils.load_checkpoint(args.model_path, model_g, None)
+
+    def infer_forward(text, text_lengths, scales, sid=None):
+        noise_scale = scales[0]
+        length_scale = scales[1]
+        noise_scale_w = scales[2]
+        audio = model_g.infer(
+            text,
+            text_lengths,
+            noise_scale=noise_scale,
+            length_scale=length_scale,
+            noise_scale_w=noise_scale_w,
+            sid=sid,
+        )[0]
+
+        return audio
+
+    model_g.forward = infer_forward
+
+    dummy_input_length = 50
+    sequences = torch.randint(
+        low=0, high=len(symbols), size=(1, dummy_input_length), dtype=torch.long
+    )
+    sequence_lengths = torch.LongTensor([sequences.size(1)])
+
+    sid: Optional[torch.LongTensor] = None
+    if hps.data.n_speakers > 1:
+        sid = torch.LongTensor([0])
+
+    # noise, length, noise_w
+    scales = torch.FloatTensor([0.667, 1.0, 0.8])
+    dummy_input = (sequences, sequence_lengths, scales, sid)
+
+    # Export
+    torch.onnx.export(
+        model=model_g,
+        args=dummy_input,
+        f=str(args.output),
+        verbose=False,
+        opset_version=OPSET_VERSION,
+        input_names=["input", "input_lengths", "scales", "sid"],
+        output_names=["output"],
+        dynamic_axes={
+            "input": {0: "batch_size", 1: "phonemes"},
+            "input_lengths": {0: "batch_size"},
+            "output": {0: "batch_size", 1: "time1", 2: "time2"},
+        },
+    )
+
+    print(f"Exported model to {args.output}")
+
+
+if __name__ == "__main__":
+    main()

filelists/final_annotation_train.txt

@@ -0,0 +1,137 @@
+./custom_character_voice/linghua/processed_0.wav|ㄉㄠˋㄑㄧˉ ㄕㄣˊㄌㄧˇ ㄌㄧㄡˊㄊㄞˋ ㄉㄠˉㄕㄨˋ ㄐㄧㄝˉ ㄔㄨㄢˊㄕㄣˊ ㄌㄧˇㄌㄧㄥˊㄏㄨㄚˊ ㄘㄢˉㄕㄤˋ。
+./custom_character_voice/linghua/processed_1.wav|ㄑㄧㄥˇ ㄉㄨㄛˉ ㄓˇㄐㄧㄠˋ ㄚ˙。
+./custom_character_voice/linghua/processed_2.wav|ㄓㄜˋㄧㄤˋ ㄧㄡˉㄒㄧㄢˊㄢˉㄨㄣˇ ㄉㄜ˙ ㄕˊㄍㄨㄤˉ, ㄖㄨˊㄍㄨㄛˇ ㄗㄞˋ ㄉㄨㄛˉ ㄧˉㄉㄧㄢˇ ㄐㄧㄡˋ ㄏㄠˇ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_3.wav|ㄨㄛˇ ㄓㄣˉ ㄊㄢˉㄒㄧㄣˉ ㄚ˙。
+./custom_character_voice/linghua/processed_4.wav|ㄐㄧㄡˋ ㄏㄜˊ ㄔㄚˊ ㄧˉㄧㄤˋ, ㄒㄧˋㄒㄧˋㄆㄧㄣˇㄨㄟˋ, ㄘㄞˊㄋㄥˊ ㄌㄧˇㄐㄧㄝˇ ㄑㄧˊㄓㄨㄥˉ ㄈㄥˉㄧㄚˇ。
+./custom_character_voice/linghua/processed_5.wav|ㄉㄡˉ ㄕˋ ㄌㄩˇㄒㄧㄥˊㄓㄜˇ ㄆㄧㄥˊㄖˋ ㄌㄧˇ ㄉㄜ˙ ㄕㄥˉㄏㄨㄛˊ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_6.wav|ㄍㄢˇㄐㄩㄝˊ ㄧㄡˋ ㄉㄨㄛˉ ㄌㄧㄠˇㄐㄧㄝˇ ㄌㄜ˙ ㄋㄧˇ ㄧˉㄒㄧㄝˉ。
+./custom_character_voice/linghua/processed_7.wav|ㄐㄧㄢˋ ㄅㄠˋㄧㄝˋㄇㄧㄥˊ ㄍㄨㄥˉㄏㄨㄞˊ ㄅㄠˇ。
+./custom_character_voice/linghua/processed_8.wav|ㄙㄨㄟˊ ㄨㄛˇ ㄧˉㄊㄨㄥˊ ㄅㄧˋㄩˇ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_9.wav|ㄙㄨㄛˇ ㄉㄚˋㄖㄣˊ, ㄕˋ ㄗㄞˋ ㄙㄨˋㄕㄨㄛˉ ㄕㄣˊㄇㄜ˙ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_10.wav|ㄐㄧˋㄧㄣˊㄓㄨㄤˉㄙㄨˋ, ㄐㄩˊㄍㄠˉㄧㄥˋ ㄑㄩㄥˊㄓˉ。
+./custom_character_voice/linghua/processed_11.wav|ㄣˊ…… ㄇㄟˇㄐㄧㄥˇ ㄉㄤˉㄑㄧㄢˊ, ㄓˇㄔㄚˋ ㄧˉㄏㄨˊ ㄔㄚˊ ㄩˇ ㄓˉ ㄒㄧㄤˉㄔㄣˋ ㄋㄜ˙。
+./custom_character_voice/linghua/processed_12.wav|ㄧㄠˋ ㄑㄩˋ ㄋㄚˇㄅㄧㄢˉ ㄗㄡˇㄗㄡˇ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_13.wav|ㄧㄢˇㄐㄧㄥˉ, ㄒㄧㄤˋ ㄓㄨˋㄈㄥˉ ㄔㄨㄟˉ ㄌㄞˊ ㄉㄜ˙ ㄈㄤˉㄒㄧㄤˋ。
+./custom_character_voice/linghua/processed_14.wav|ㄞˉㄧㄚˉ, ㄏㄣˇ ㄕㄨˉㄈㄨˊ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_15.wav|ㄌㄩˇㄒㄧㄥˊㄓㄜˇ。
+./custom_character_voice/linghua/processed_16.wav|ㄓㄜˋㄧㄤˋ ㄗㄞˋ ㄑㄧㄥˉㄔㄣˊ ㄐㄧㄢˋ ㄋㄧˇ ㄧˊㄇㄧㄢˋ, ㄨㄛˇㄏㄨㄟˋ ㄖㄣˇㄅㄨˊㄓㄨˋ ㄐㄩㄝˊㄉㄜˊ, ㄐㄧㄝˉㄒㄧㄚˋ ㄌㄞˊㄐㄧㄤˉ ㄕˋ ㄕㄨㄣˋㄌㄧˋ ㄉㄜ˙ ㄧˉㄊㄧㄢˉ。
+./custom_character_voice/linghua/processed_18.wav|ㄔㄚˊㄈㄢˋ ㄓˉㄏㄡˋ, ㄋㄢˊㄇㄧㄢˇ ㄌㄩㄝˋㄧㄡˇ ㄎㄨㄣˋㄐㄩㄢˋ ㄕˋㄈㄡˇ ㄧㄡˇ ㄒㄧㄥˋㄓˋ ㄒㄧㄚˋㄆㄢˊ ㄑㄧˊ ㄊㄧˊㄕㄣˊ ㄋㄜ˙?
+./custom_character_voice/linghua/processed_20.wav|ㄏㄨㄟˋ ㄕˋ ㄧˊㄍㄜˋ ㄌㄧㄤˊㄒㄧㄠˉ ㄋㄜ˙。
+./custom_character_voice/linghua/processed_21.wav|ㄓˉㄕˋ ㄇㄥˋ  ㄏㄜˊㄒㄩˉ ㄒㄧㄥˇ。
+./custom_character_voice/linghua/processed_22.wav|ㄅㄨˋㄅㄧˇㄓㄣˉ ㄖㄨˊ, ㄧˉㄒㄧㄤˉㄏㄨㄟˋ。
+./custom_character_voice/linghua/processed_23.wav|ㄉㄠˋㄑㄧˉ ㄇㄨˋㄈㄨˇ ㄕㄜˋ ㄈㄥˋㄒㄧㄥˊ ㄕㄣˊㄌㄧˇㄐㄧㄚˉ, ㄨㄟˋ ㄩˊ ㄉㄠˋㄑㄧˉ ㄇㄧㄥˊㄇㄣˊ ㄓㄨㄥˉ ㄉㄜ˙ ㄅㄧˇㄊㄡˊ ㄓˉ ㄍㄜˊㄨㄟˋ。
+./custom_character_voice/linghua/processed_24.wav|ㄗㄨㄛˋㄨㄟˊ ㄙㄢˉ ㄈㄥˋㄒㄧㄥˊ ㄓˉㄧˉ, ㄓㄤˇㄍㄨㄢˇ ㄐㄧˋㄙˋ ㄏㄨㄛˊㄉㄨㄥˋ ㄩˇ ㄖㄣˊㄨㄣˊ ㄧˋㄕㄨˋ。
+./custom_character_voice/linghua/processed_25.wav|ㄕㄨㄤˉㄑㄧㄣˉ ㄍㄨㄛˋㄕˋ ㄓˉㄏㄡˋ, ㄗㄨˊ ㄋㄟˋ ㄉㄜ˙ ㄉㄚˋㄒㄧㄠˇ ㄕˋㄨˋ ㄅㄧㄢˋ ㄧㄡˊ ㄒㄩㄥˉㄓㄤˇ ㄏㄜˊ ㄨㄛˇ ㄔㄥˊㄉㄢˉ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_26.wav|ㄏㄣˇㄉㄨㄛˉ ㄖㄣˊㄧㄣˉ ㄨㄟˋ ㄨㄛˇ ㄕˋ ㄅㄞˊㄌㄨˋ ㄍㄨㄥˉㄓㄨˇ, ㄕˋ ㄕㄜˋ ㄈㄥˋㄒㄧㄥˊ ㄕㄣˊㄌㄧˇㄐㄧㄚˉ ㄉㄜ˙ ㄉㄚˋ ㄒㄧㄠˇㄐㄧㄝˇ, ㄦˊ ㄐㄧㄥˋㄓㄨㄥˋ ㄨㄛˇ。
+./custom_character_voice/linghua/processed_27.wav|ㄊㄚˉㄇㄣ˙ ㄙㄨㄛˇ ㄐㄧㄥˋㄓㄨㄥˋ ㄉㄜ˙, ㄓˇㄕˋ ㄨㄛˇ ㄙㄨㄛˇ ㄕㄣˉㄔㄨˋ ㄉㄜ˙ ㄉㄧˋㄨㄟˋ, ㄩˇㄌㄧㄥˊㄏㄨㄚˊ ㄨㄛˇ ㄕˋ ㄗㄣˇㄧㄤˋ ㄉㄜ˙ ㄖㄣˊ ㄅㄧㄥˋ ㄨˊ ㄍㄨㄢˉㄒㄧˋ。
+./custom_character_voice/linghua/processed_28.wav|ㄨㄛˇ ㄒㄧㄤˇ, ㄋㄥˊ ㄓㄣˉㄓㄥˋ ㄗㄡˇㄐㄧㄣˋ ㄨㄛˇ ㄉㄜ˙, ㄏㄨㄛˋㄒㄩˇ ㄓˇㄧㄡˇ, ㄖㄨˊㄐㄧㄣˉ ㄉㄜ˙ ㄨㄛˇ。
+./custom_character_voice/linghua/processed_29.wav|ㄧˉㄐㄧㄡˋ ㄒㄧㄤˇ ㄔㄥˊㄨㄟˋ ㄓˊㄉㄜ˙ ㄉㄚˋㄐㄧㄚˉ ㄒㄧㄣˋㄖㄣˋ ㄉㄜ˙ ㄖㄣˊ。
+./custom_character_voice/linghua/processed_30.wav|ㄍㄨˇㄨˇ ㄨㄛˇ ㄉㄜ˙ ㄩㄢˊㄧㄣˉ, ㄧˇ ㄅㄨˋㄗㄞˋ ㄕˋ ㄐㄧㄢˉㄕㄤˋ ㄉㄜ˙ ㄗㄜˊㄖㄣˋ, ㄏㄨㄛˋ ㄊㄚˉㄖㄣˊ ㄉㄜ˙ ㄑㄧˉㄉㄞˋ。
+./custom_character_voice/linghua/processed_31.wav|ㄕˋ ㄧㄣˉㄨㄟˋ…… ㄋㄧˇ ㄧㄝˇ ㄕˋ ㄓㄜˋㄧㄤˋ ㄉㄜ˙ ㄖㄣˊ ㄚ˙。
+./custom_character_voice/linghua/processed_32.wav|ㄖㄨˊㄍㄨㄛˇ ㄋㄧㄣˊ ㄧㄡˇㄎㄨㄥˋ, ㄨㄛˇㄇㄣ˙ ㄧˉㄅㄨˋ ㄇㄨˋㄌㄨˋ ㄔㄚˊㄕˋ ㄖㄨˊㄏㄜˊ?
+./custom_character_voice/linghua/processed_33.wav|ㄗㄞˋ ㄓㄜˋㄧㄤˋ ㄊㄧㄢˊㄐㄧㄥˋ ㄉㄜ˙ ㄖˋㄗ˙, ㄌㄩㄝˋㄐㄧㄚˉ ㄐㄧㄠˉㄌㄧㄡˊ ㄔㄚˊㄧˋ ㄒㄧㄣˉㄉㄜˊ, ㄒㄧㄤˇㄌㄞˊ ㄕˋ ㄆㄛˇㄐㄩˋ ㄧㄚˇㄑㄩˋ ㄉㄜ˙。
+./custom_character_voice/linghua/processed_34.wav|ㄖㄨˊㄍㄨㄛˇ ㄧㄡˇ ㄐㄧˉㄏㄨㄟˋ ㄉㄜ˙ㄏㄨㄚˋ, ㄨㄛˇㄒㄧㄤˇㄕˋ ㄓㄜ˙ ㄏㄜˊ ㄋㄧˇ ㄍㄨㄥˋㄉㄨˋ ㄧˋㄍㄨㄛˊ ㄉㄜ˙ ㄐㄧㄝˊㄖˋ。
+./custom_character_voice/linghua/processed_35.wav|ㄗㄨㄣˉㄒㄩㄣˊ ㄉㄜ˙ ㄉㄤˉㄉㄧˋ ㄈㄥˉㄙㄨˊ， ㄌㄧˇㄧˊ ㄍㄨㄟˉㄈㄢˋ, ㄏㄞˊㄧㄡˇ ㄅㄢˋㄕㄡˇ ㄌㄧˇ ㄉㄜ˙ ㄊㄨㄟˉㄐㄧㄢˋ。
+./custom_character_voice/linghua/processed_36.wav|ㄎㄜˇㄧˇ ㄇㄚˊㄈㄢˊ ㄋㄧˇ… ㄧˉㄧˉ ㄓˇㄉㄠˇ ㄨㄛˇ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_37.wav|ㄕㄣˊㄓˉㄧㄢˇ, ㄐㄧˊ ㄕˋ ㄒㄩㄥˉㄏㄨㄞˊㄉㄚˋㄓˋ ㄓˉㄖㄣˊ ㄙㄨㄛˇ ㄏㄨㄛˋ ㄉㄜ˙ ㄧㄥˉㄕㄡˋ。
+./custom_character_voice/linghua/processed_38.wav|ㄖㄨˊㄍㄨㄛˇ ㄨㄣˋ ㄨㄛˇ ㄧㄡˇ ㄕㄣˊㄇㄜ˙ ㄓˋㄒㄧㄤˋ ㄉㄜ˙ㄏㄨㄚˋ。
+./custom_character_voice/linghua/processed_39.wav|ㄓㄜˋㄍㄜˋ ㄏㄞˊㄕˋ ㄅㄠˇㄇㄧˋ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_40.wav|ㄓˇㄕˋ ㄧˊㄍㄜˋ ㄨㄟˉㄅㄨˋㄗㄨˊㄉㄠˋ ㄉㄜ˙ ㄇㄥˋㄒㄧㄤˇ ㄅㄚˋㄌㄜ˙。
+./custom_character_voice/linghua/processed_41.wav|ㄔㄤˊㄕㄨㄛˉ  ㄔㄢˊ ㄔㄚˊ ㄧˉㄨㄟˋ。
+./custom_character_voice/linghua/processed_42.wav|ㄐㄧㄢˋ ㄔㄢˊ ㄧˋ ㄖㄨˊ。
+./custom_character_voice/linghua/processed_43.wav|ㄋㄚˋㄇㄜ˙ ㄐㄧㄢˋ ㄏㄜˊ ㄔㄚˊ, ㄧㄡˋ ㄕˋ ㄕㄣˊㄇㄜ˙ ㄍㄨㄢˉㄒㄧˋ ㄋㄜ˙?
+./custom_character_voice/linghua/processed_44.wav|ㄋㄧˇ ㄗㄞˋ ㄔㄥˊㄓㄨㄥˉ, ㄐㄧㄢˋㄍㄨㄛˋ ㄎㄨˉㄨㄟˇ ㄉㄜ˙ ㄧㄥˉㄏㄨㄚˉㄕㄨˋ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_45.wav|ㄎㄨˉㄓˉ ㄇㄟˇ  ㄖㄤˋ ㄨㄛˇ ㄒㄧㄤˇㄉㄠˋ ㄔㄨㄣˉㄊㄧㄢˉ ㄕㄥˋㄎㄞˉ ㄓˉㄐㄧㄥˇ。
+./custom_character_voice/linghua/processed_46.wav|ㄅㄨˋㄍㄨㄛˋ, ㄅㄧㄝˊㄖㄣˊ ㄙˋㄏㄨˉ ㄅㄧㄥˋ ㄅㄨˋ ㄓㄜˋㄇㄜ˙ ㄒㄧㄤˇ。
+./custom_character_voice/linghua/processed_47.wav|ㄎㄞˉㄏㄨㄚˉ ㄉㄜ˙ ㄧㄣˉㄕㄨˋㄏㄨㄟˋ ㄅㄟˋ ㄧˊㄗㄡˇ。
+./custom_character_voice/linghua/processed_48.wav|ㄐㄧㄡˋㄙㄨㄢˋ ㄧˉㄘˋ ㄧㄝˇㄏㄠˇ, ㄓㄣˉㄒㄧㄤˇ ㄎㄢˋㄉㄠˋ ㄊㄚˉ ㄗㄞˋㄘˋ ㄎㄞˉㄈㄤˋ。
+./custom_character_voice/linghua/processed_49.wav|ㄕˋ ㄨㄛˇ ㄏㄣˇ ㄓㄨㄥˋㄧㄠˋ ㄉㄜ˙ ㄆㄥˊㄧㄡˇ。
+./custom_character_voice/linghua/processed_50.wav|ㄊㄧㄢˉㄌㄥˇ ㄏㄜˊ ㄧㄤˊㄍㄨㄤˉ, ㄗㄨㄥˇㄕˋ ㄍㄢˇㄖㄢˇ ㄓㄜ˙ ㄨㄛˇ。
+./custom_character_voice/linghua/processed_51.wav|ㄧˋㄧˋ ㄕㄤˋ ㄌㄞˊㄕㄨㄛˉ, ㄊㄚˉ ㄐㄧㄡˋ ㄒㄧㄤˋㄕˋ ㄨㄛˇ ㄉㄜ˙ ㄌㄧㄥˋ ㄧˊㄍㄜˋ ㄒㄩㄥˉㄓㄤˇ ㄧˉㄧㄤˋ。
+./custom_character_voice/linghua/processed_52.wav|ㄔㄥˊㄨㄟˋ ㄌㄜ˙ ㄕㄣˊㄌㄧˋ ㄐㄧㄚˉ ㄉㄜ˙ ㄧˉㄩㄢˊ。
+./custom_character_voice/linghua/processed_53.wav|ㄌㄧㄥˇ ㄈㄥˋㄒㄧㄥˊ ㄍㄨㄥˉㄗㄨㄛˋ ㄉㄜ˙ ㄐㄧㄡˇㄊㄧㄠˊ ㄕㄚˉㄌㄨㄛˊ, ㄊㄚˉ ㄗㄨㄥˇㄕˋ ㄧˉㄌㄧㄢˇ ㄧㄢˊㄙㄨˋ。
+./custom_character_voice/linghua/processed_54.wav|ㄊㄚˉ ㄘㄥˊㄐㄧㄥˉ ㄑㄧˇㄍㄨㄛˋ ㄐㄧˇㄘˋ ㄓㄥˉㄉㄨㄢˉ。
+./custom_character_voice/linghua/processed_55.wav|ㄊㄚˉ ㄅㄣˇㄓˋ ㄕˋ ㄓㄨㄥˉㄧˋ ㄓˉㄕˋ, ㄓㄜˋㄉㄧㄢˇ ㄨˊㄎㄜˇㄈㄡˇㄖㄣˋ。
+./custom_character_voice/linghua/processed_56.wav|ㄙㄨˉㄧㄝˇ ㄩㄝˋㄒㄧㄢˉㄕㄥˉ。
+./custom_character_voice/linghua/processed_57.wav|ㄊㄚˉ ㄏㄜˊ ㄐㄧㄡˇㄊㄧㄠˊ ㄒㄧㄠˇㄐㄧㄝˇ ㄧˉㄧㄤˋ, ㄕˋ ㄊㄧㄢˉㄌㄧㄥˇ ㄈㄥˋㄒㄧㄥˊ ㄉㄜ˙ ㄖㄣˊ。
+./custom_character_voice/linghua/processed_58.wav|ㄊㄚˉ…… ㄧㄝˇ ㄕˋ ㄧˊㄍㄜˋ ㄏㄣˇ ㄧㄡˇ ㄩㄢˊㄗㄜˊ ㄉㄜ˙ ㄖㄣˊ。
+./custom_character_voice/linghua/processed_59.wav|ㄓㄜˋㄒㄧㄝˉ ㄩㄢˊㄗㄜˊ ㄉㄜ˙ ㄐㄧㄢˉㄔˊ, ㄕㄣˋㄓˋ ㄅㄧˇ ㄐㄧㄡˇㄊㄧㄠˊ ㄒㄧㄠˇㄐㄧㄝˇ ㄍㄥˋ ㄓˊㄓㄨㄛˊ。
+./custom_character_voice/linghua/processed_60.wav|ㄅㄨˋㄍㄨㄛˋ, ㄕㄣˊㄇㄜ˙ ㄕˋ ㄧㄥˉㄍㄞˉ ㄅㄟˋ ㄙㄨㄢˋ ㄗㄞˋ ㄓㄜˋㄒㄧㄝˉ ㄩㄢˊㄗㄜˊ ㄓˉㄋㄟˋ…… ㄨㄛˇ ㄒㄧㄤˇ, ㄏㄨㄛˋㄒㄩˇ ㄓˇㄧㄡˇ ㄌㄨˋ ㄧㄝˇ ㄩㄢ�� ㄒㄧㄢˉㄕㄥˉ ㄗˋㄐㄧˇ ㄓˉㄉㄠˋ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_61.wav|ㄒㄧㄠˇ ㄧㄡˋㄓˋ ㄏㄞˊㄗ˙, ㄗㄨㄟˋㄐㄧㄣˋ ㄧㄡˇ ㄇㄟˊ ㄧㄡˇㄍㄟˇ ㄋㄧˇ ㄊㄧㄢˉㄕㄣˊㄇㄜ˙ ㄇㄚˊㄈㄢˊ ㄋㄜ˙?
+./custom_character_voice/linghua/processed_62.wav|ㄖㄨˊㄍㄨㄛˇ ㄎㄢˋㄐㄧㄢˋ ㄊㄚˉ ㄊㄡˉ ㄌㄢˇ, ㄎㄜˇㄧˇ ㄓˊㄐㄧㄝˉ ㄍㄠˋㄙㄨˋ ㄨㄛˇ。
+./custom_character_voice/linghua/processed_63.wav|ㄌㄧˇㄙㄨㄛˇ ㄉㄤˉㄖㄢˊ ㄉㄜ˙ ㄎㄢˋㄈㄚˇ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_64.wav|ㄅㄨˋㄍㄞˉ ㄧㄡˊ ㄨㄛˇ ㄉㄥˇ ㄒㄧㄚˋㄕㄨˇ ㄙㄨㄟˊㄧˋ ㄧˋㄌㄨㄣˋ。
+./custom_character_voice/linghua/processed_65.wav|ㄏㄥˉ, ㄐㄧㄤˉㄐㄩㄣˉ ㄉㄚˋㄖㄣˊ ㄊㄚˉ ㄗㄞˋ ㄓㄨㄟˉㄑㄧㄡˊ ㄩㄥˇㄏㄥˊ ㄓˉ ㄌㄨˋㄕㄤˋ, ㄎㄜˇㄋㄥˊ ㄧㄝˇ ㄏㄣˇ ㄍㄨˉㄉㄨˊ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_66.wav|ㄧˉㄉㄠˉ, ㄅㄧㄥˋㄑㄧㄝˇ ㄏㄨㄛˊ ㄌㄜ˙ ㄒㄧㄚˋㄌㄞˊ。 ㄍㄞˉ ㄕㄨㄛˉ ㄕˋ ㄎㄢˉㄔㄥˉ ㄨㄟˇㄧㄝˋ ㄉㄜ˙ ㄐㄧㄥˉㄌㄧˋ ㄌㄜ˙ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_67.wav|ㄙㄨㄟˉㄖㄢˊ ㄉㄨㄟˋ ㄨㄛˇ ㄌㄞˊㄕㄨㄛˉ, ㄊㄚˉ ㄕˋ ㄓㄣˉㄓㄥˋ ㄉㄜ˙ ㄕㄣˊㄇㄧㄥˊ。
+./custom_character_voice/linghua/processed_68.wav|ㄎㄜˇㄧˇ ㄍㄥˉㄍㄞˇ ㄉㄠˋㄑㄧˉ ㄉㄜ˙ ㄇㄧㄥˋㄩㄣˋ。
+./custom_character_voice/linghua/processed_69.wav|ㄉㄢˋㄕˋ ㄖㄨˊㄍㄨㄛˇ ㄕˋ ㄏㄜˊ ㄋㄧˇ ㄑㄧˇ ㄌㄜ˙ ㄔㄨㄥˉㄊㄨˉ ㄉㄜ˙ㄏㄨㄚˋ, ㄨㄛˇ ㄧˊㄉㄧㄥˋ ㄏㄨㄟˋ ㄓㄢˋ ㄗㄞˋ ㄋㄧˇ ㄓㄜˋ ㄧˉㄅㄧㄢˉ ㄉㄜ˙。
+./custom_character_voice/linghua/processed_70.wav|ㄅㄚˉㄓㄨㄥˋ ㄍㄨㄥˉㄙˉ ㄉㄚˋㄖㄣˊ ㄉㄜ˙ ㄏㄜˊㄗㄨㄛˋ ㄒㄧㄤˋㄌㄞˊ ㄏㄣˇ ㄩˊㄎㄨㄞˋ。
+./custom_character_voice/linghua/processed_71.wav|ㄎㄢˋ, ㄘㄠˉㄅㄢˋ ㄐㄧㄝˊㄑㄧㄥˋㄑㄧㄥˋㄉㄧㄢˇ ㄈㄟˉㄔㄤˊ ㄌㄠˊㄕㄣˊㄈㄟˋㄌㄧˋ, ㄧㄥˊㄕㄡˉ ㄉㄨㄛˉㄅㄢˋ ㄧㄝˇ ㄅㄨˋ ㄏㄠˇㄎㄢˋ。
+./custom_character_voice/linghua/processed_72.wav|ㄅㄚˉㄓㄨㄥˋ ㄍㄨㄥˉㄙˉ ㄉㄚˋㄖㄣˊ ㄘㄠˉㄅㄢˋ ㄉㄜ˙ ㄔㄢˇㄧㄝˋ, ㄓㄣˉㄉㄜ˙ ㄐㄧˋ ㄈㄥˉㄧㄚˇ ㄧㄡˋ ㄧㄡˇ ㄕㄡˉㄔㄥˊ。
+./custom_character_voice/linghua/processed_73.wav|ㄗㄨㄛˋㄨㄟˋ ㄕㄣˊㄌㄧˋ ㄐㄧㄚˉ ㄉㄜ˙ ㄐㄧㄚˉㄓㄨˇ, ㄒㄩㄥˉㄓㄤˇ ㄆㄧㄥˊㄖˋ ㄒㄩˉㄧㄠˋ ㄔㄨˉㄇㄧㄢˋ ㄓㄨˉㄉㄨㄛˉㄔㄤˇ ㄏㄜˊ。
+./custom_character_voice/linghua/processed_74.wav|ㄨㄛˇ ㄙㄨㄟˉ ㄐㄧㄣˇㄌㄧˋ ㄈㄣˉㄉㄢˉ ㄒㄩㄥˉㄓㄤˇ ㄐㄧㄢˉㄕㄤˋ ㄙㄨㄛˇ ㄈㄨˋㄉㄢˉ ㄉㄜ˙ ㄗㄜˊㄖㄣˋ, ㄑㄩㄝˋ ㄧㄝˇ ㄨˊㄈㄚˇㄏㄨㄢˇ ㄏㄜˊ ㄊㄚˉ ㄔㄤˊㄋㄧㄢˊ ㄐㄧˉㄧㄚˉ ㄗㄞˋ ㄕㄣˉ ㄉㄜ˙ ㄆㄧˊㄐㄩㄢˋ。
+./custom_character_voice/linghua/processed_75.wav|ㄨㄛˇ ㄧㄝˇ ㄕˋ ㄉㄤˉㄕˊ ㄨㄟˋㄌㄜ˙ ㄇㄟˋㄇㄟˋ ㄓㄨㄛˊㄒㄧㄤˇ, ㄒㄧˉㄨㄤˋ ㄋㄧˇ ㄋㄥˊ ㄑㄩㄢˋㄧㄢˊ, ㄖㄤˋ ㄒㄩㄥˉㄓㄤˇ ㄉㄨㄛˉㄉㄨㄛˉ ㄓㄨˋㄧˋ ㄕㄣˉㄊㄧˇ ㄚ˙。
+./custom_character_voice/linghua/processed_76.wav|ㄐㄧㄝˊㄑㄧㄥˋㄑㄧㄥˋㄉㄧㄢˇ ㄕˋ ㄕㄥˋ ㄈㄥˋㄒㄧㄥˊ ㄏㄜˊ ㄨˉㄋㄩˇ ㄓㄨㄥˉ ㄉㄜ˙ ㄗㄜˊㄖㄣˋ。
+./custom_character_voice/linghua/processed_77.wav|ㄗㄨㄛˋㄨㄟˋ ㄧㄢˉㄏㄨㄛˇ ㄓㄨㄢˉㄐㄧㄚˉ, ㄧㄝˇ ㄉㄜ˙ ㄑㄩㄝˋㄋㄥˊ ㄖㄤˋ ㄑㄧˋㄈㄣˉ ㄖㄜˋㄌㄧㄝˋ ㄑㄧˇㄌㄞˊ。
+./custom_character_voice/linghua/processed_78.wav|ㄏㄜˊㄗㄨㄛˋ ㄉㄨㄛˉ ㄌㄜ˙, ㄧㄣˉㄦˊ ㄐㄧㄢˋㄐㄧㄢˋ ㄕㄨˊㄌㄨㄛˋ。
+./custom_character_voice/linghua/processed_79.wav|ㄓˋㄢˉ ㄏㄜˊ ㄒㄧㄠˉㄈㄤˊ ㄨㄣˋㄊㄧˊ ㄇㄚ˙? ㄨㄛˇㄇㄣ˙ ㄧㄝˇ ㄏㄨㄟˋ ㄧˉ ㄅㄧㄥˋ ㄋㄚˋㄖㄨˋ ㄩˋㄒㄧㄢˉ ㄍㄨㄟˉㄏㄨㄚˋ ㄉㄜ˙。
+./custom_character_voice/linghua/processed_80.wav|ㄓㄜˋㄒㄧㄝˉ ㄩㄢˊㄧㄣˉ ㄦˊ ㄎㄢˋㄅㄨˊㄉㄠˋ ㄧㄥˉㄏㄨㄚˉ, ㄘㄞˊ ㄏㄨㄟˋ ㄖㄤˋ ㄖㄣˊㄇㄣ˙ ㄒㄧㄣˉㄓㄨㄥˉ ㄌㄧㄡˊㄒㄧㄚˋ ㄑㄩㄝˉㄏㄢˋ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_81.wav|ㄋㄧˇ ㄉㄜ˙ ㄑㄧㄥˇㄑㄧㄡˊ, ㄉㄨㄟˋ ㄨㄛˇ ㄌㄞˊㄕㄨㄛˉ ㄏㄣˇ ㄊㄜˋㄅㄧㄝˊ ㄋㄜ˙ ㄐㄧˋㄖㄢˊ ㄅㄚˇ ㄋㄧˇ ㄉㄤˋㄗㄨㄛˋ ㄆㄥˊㄧㄡˇ, ㄨㄛˇ ㄧㄝˇ ㄧㄥˉ ㄊㄢˇㄔㄥˊㄧˇㄉㄞˋ。
+./custom_character_voice/linghua/processed_82.wav|ㄅㄨˋㄍㄨㄛˋ, ㄕˋㄍㄨㄢˉ ㄕㄣˊㄌㄧˇㄐㄧㄚˉ ㄉㄜ˙ ㄇㄧˋㄇㄧˋ, ㄏㄞˊ ㄒㄧˉㄨㄤˋ ㄋㄧˇ ㄋㄥˊ ㄕㄡˇㄎㄡˇㄖㄨˊㄆㄧㄥˊ。
+./custom_character_voice/linghua/processed_83.wav|ㄕㄣˊㄌㄧˋ ㄐㄧㄚˉㄧㄣˉ ㄨㄟˋ ㄇㄟˊㄋㄥˊ ㄅㄠˇㄏㄨˋ ㄏㄠˇ ㄍㄨㄛˊㄅ���ˇㄐㄧˊ ㄅㄧㄝˊ ㄉㄜ˙ ㄉㄠˉㄍㄨㄥˉ, ㄗㄠˉㄕㄡˋ ㄌㄜ˙ ㄅㄨˋㄒㄧㄠˇ ㄉㄜ˙ ㄔㄨㄥˉㄐㄧˉ。
+./custom_character_voice/linghua/processed_84.wav|ㄅㄧㄝˊㄖㄣˊ ㄓㄨㄥˉ ㄧㄣˉㄇㄡˊ ㄙㄨㄢˋㄐㄧˋ ㄨㄛˇㄇㄣ˙ ㄕㄜˊㄙㄨㄣˇ ㄌㄜ˙ ㄓㄨˉㄉㄨㄛˉ ㄔㄣˊ ㄒㄧㄚˋ ㄕㄡˋㄉㄠˋ ㄒㄩˇㄉㄨㄛˉ ㄗㄜˊㄈㄚˊ。
+./custom_character_voice/linghua/processed_85.wav|ㄕㄣˋㄓˋ ㄧㄣˉ ㄓˉ ㄗㄠˇㄕㄨㄞˉ ㄕㄣˊㄌㄧˋ ㄐㄧㄚˉ ㄗㄞˋ ㄇㄨˋㄈㄨˇ ㄓㄨㄥˉ ㄉㄜ˙ ㄉㄧˋㄨㄟˋ ㄧㄝˇ ㄧˉㄌㄨㄛˋㄑㄧㄢˉㄓㄤˋ。
+./custom_character_voice/linghua/processed_86.wav|ㄏㄠˇ ㄗㄞˋ ㄒㄩㄥˉㄓㄤˇ ㄐㄧˋㄖㄣˋㄏㄡˋㄌㄧˋ ㄨㄢˇ ㄎㄨㄤˊㄌㄢˊ。
+./custom_character_voice/linghua/processed_87.wav|ㄐㄧㄚˉㄇㄣˊ ㄙㄨㄟˉ ㄧˇ ㄈㄨˋ ㄒㄧㄥˋ, ㄕㄜˋㄈㄥˉㄒㄧㄥˇ ㄧˉㄒㄧˉ ㄧㄝˇ ㄕㄤˋㄒㄧㄚˋ ㄑㄧˊㄒㄧㄣˉ, ㄉㄢˋ ㄉㄚˋㄕˋ ㄧㄠˋㄕˋ ㄈㄤˉㄇㄧㄢˋ ㄖㄥˊㄎㄠˋ ㄒㄩㄥˉㄓㄤˇ ㄉㄧㄥˋㄉㄨㄛˊ。
+./custom_character_voice/linghua/processed_88.wav|ㄊㄚˉㄇㄣ˙ ㄙㄨㄛˇㄔㄨㄢˊ ㄉㄜ˙ ㄉㄨㄢˋ ㄉㄠˉ ㄓˉㄕㄨˋ, ㄧㄝˇ ㄧㄣˉ ㄒㄧㄥˉㄒㄧㄤˋ， ㄩㄥˋㄊㄨˊ， ㄎㄨㄤˋㄓˊ， ㄌㄨˊㄏㄨㄛˇ ㄏㄨㄢˊㄐㄧㄥˋ， ㄖㄣˊ ㄓˉ ㄒㄧㄥˋㄍㄜˊ， ㄩㄢˊㄙㄨˋ ㄅㄧㄢˋㄏㄨㄚˋ ㄉㄜ˙ ㄅㄨˋㄊㄨㄥˊ ㄦˊ ㄧㄡˇㄙㄨㄛˇ ㄑㄩˉㄈㄣˉ。
+./custom_character_voice/linghua/processed_89.wav|ㄕˋ ㄉㄠˉ ㄍㄨㄥˉ ㄓˉㄐㄧㄢˉ ㄙㄨㄛˇㄕㄨㄛˉ ㄉㄜ˙ ㄌㄟˊㄉㄧㄢˋ ㄨˇㄔㄨㄢˉ。
+./custom_character_voice/linghua/processed_90.wav|ㄘㄤˊㄇㄧㄥˊ ㄉㄠˉ ㄉㄜ˙ ㄉㄠˉㄍㄨㄥˉ, ㄧㄝˇ ㄅㄟˋ ㄙㄨㄢˋㄗㄨㄛˋ ㄕˋ ㄅㄣˇㄌㄧㄥˇ ㄊㄨㄥˉㄕㄣˊ ㄉㄜ˙ ㄕㄣˊㄕˋ ㄒㄧㄤˉㄍㄨㄢˉ ㄖㄣˊㄩㄢˊ, ㄍㄨㄟˉㄕㄨˇ ㄊㄨㄥˇㄔㄡˊ ㄨㄣˊㄏㄨㄚˋ， ㄧˋㄕㄨˋ， ㄐㄧˋㄙˋ ㄉㄜ˙ ㄕㄜˋㄈㄥˋ ㄒㄧㄥˊㄧˋ ㄆㄞˋ ㄍㄨㄢˇㄌㄧˇ。
+./custom_character_voice/linghua/processed_91.wav|ㄔㄨˉㄒㄧㄢˋ ㄌㄜ˙ ㄉㄠˉㄍㄨㄥˉ ㄅㄟˋㄆㄢˋ ㄉㄜ˙ ㄕˋㄑㄧㄥˊ, ㄗˋㄖㄢˊ ㄐㄧㄡˋㄕˋ ㄕㄣˊㄌㄧˇㄐㄧㄚˉ ㄉㄜ˙ ㄉㄨˊㄅㄢˋ ㄅㄨˋㄌㄧˋ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_92.wav|ㄉㄨㄟˋ ㄨㄛˇ ㄌㄞˊ ㄕㄨㄛˉ, ㄇㄨˇㄑㄧㄣˉ ㄕˋ ㄧˋㄧˋ ㄈㄟˉㄈㄢˊ ㄉㄜ˙ ㄘㄨㄣˊㄗㄞˋ。
+./custom_character_voice/linghua/processed_93.wav|ㄈㄨˊㄓㄨㄤˉ, ㄧㄡˉㄧㄚˇ, ㄨˊㄌㄨㄣˋ ㄩˋㄉㄠˋ ㄗㄣˇㄧㄤˋ ㄉㄜ˙ ㄐㄩˊㄇㄧㄢˋ, ㄉㄡˉ ㄏㄨㄟˋ ㄌㄨˋㄔㄨˉ ㄔㄣˊㄐㄧㄣˋ ㄉㄜ˙ ㄒㄧㄠˋㄖㄨㄥˊ, ㄧˇ ㄘㄨㄥˊㄖㄨㄥˊㄅㄨˋㄆㄛˋ ㄉㄜ˙ ㄊㄞˋㄉㄨˋ, ㄘㄠˉㄔˊ ㄓㄜ˙ ㄕㄣˇㄌㄧˇ ㄐㄧㄚˉ ㄉㄚˋㄉㄚˋㄒㄧㄠˇㄒㄧㄠˇ ㄉㄜ˙ ㄕˋㄨˋ。
+./custom_character_voice/linghua/processed_94.wav|ㄍㄢˇㄑㄧㄥˊ ㄕˋ ㄨㄢˊㄇㄟˇ ㄉㄜ˙ ㄏㄨㄚˋㄕㄣˉ ㄧㄝˇ ㄅㄨˋ ㄨㄟˋㄍㄨㄛˋ。
+./custom_character_voice/linghua/processed_95.wav|ㄉㄢˋ ㄗˋㄘㄨㄥˊ ㄊㄚˉ ㄌㄧˊㄕˋ ㄉㄜ˙ ㄋㄚˋ ㄧˉㄎㄜˋㄑㄧˇ, ㄨㄛˇ ㄐㄧㄡˋ ㄕㄣˉㄑㄧㄝˋ ㄉㄧˋ ㄧˋㄕˊ ㄉㄠˋ, ㄨㄛˇ ㄧˇㄐㄧㄥˉ ㄅㄨˊㄕˋ ㄋㄚˋㄍㄜˋ ㄎㄜˇㄧˇ ㄉㄨㄛˇ ㄗㄞˋ ㄇㄨˇㄑㄧㄣˉ ㄕㄣˉㄏㄡˋ ㄉㄜ˙ ㄒㄧㄠˇㄌㄧㄥˊㄏㄨㄚˉ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_96.wav|ㄩㄢˊㄌㄞˊ ㄧㄠˋㄕㄨㄛˉ ㄉㄜ˙ㄏㄨㄚˋ, ㄎㄜˇㄋㄥˊ ㄅㄨˋㄊㄞˋ ㄈㄨˊㄏㄜˊ ㄉㄠˋㄑㄧㄝˋ ㄇㄨˋㄈㄨˇ ㄕㄜˋ ㄈㄥˉㄒㄧㄥˊ ㄕㄣˊㄌㄧˋ ㄐㄧㄚˉ ㄉㄜ˙ ㄕㄣˉㄈㄣˋ。
+./custom_character_voice/linghua/processed_97.wav|ㄅㄨˋㄍㄨㄛˋ, ㄐㄧㄡˋ ㄨㄛˇ ㄉㄜ˙ ㄆㄢˋㄉㄨㄢˋ, ㄋㄧˇ ㄧㄥˉㄍㄞˉ ㄋㄥˊ ㄖㄨㄥˊㄒㄩˇ ㄨㄛˇ ㄒㄧㄠˇㄒㄧㄠˇㄉㄜ˙ ㄖㄣˋㄒㄧㄥˋ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_98.wav|ㄕㄠˉㄨㄟˉ ㄧㄡˇㄉㄧㄢˇㄌㄟˋ ㄌㄜ˙ ㄎㄜˇㄧˇ ㄖㄤˋ ㄨㄛˇ ㄎㄠˋ ㄧˉㄒㄧㄚˋ ㄋㄧˇ ㄉㄜ˙ ㄐㄧㄢˉㄅㄤˇ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_99.wav|ㄐㄧㄡˋㄐㄧㄡˋ。
+./custom_character_voice/linghua/processed_100.wav|ㄨㄛˇ ㄉㄡˉ ㄏㄣˇ ㄒㄧˇㄏㄨㄢˉ ㄧㄚˇㄩㄝˋ ㄕˉㄘˊ ㄑㄧˊㄧˋ ㄩˇ ㄨˇㄉㄠˇ ㄋㄧˇ ㄧㄝˇ ㄍㄢˇㄒㄧㄥˋㄑㄩˋ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_102.wav|…… ㄕˋ ㄚ˙, ㄓˇㄧㄠˋ ㄕㄥˉㄏㄨㄛˊ ㄗㄞˋ ㄓㄜˋㄍㄜˋ ㄕˋㄐㄧㄝˋ ㄕㄤˋ, ㄐㄧㄡˋ ㄋㄢˊㄇㄧㄢˇ ㄩˋㄉㄠˋ ㄓㄨㄥˇㄓㄨㄥˇ ㄅㄨˋㄖㄨˊㄧˋ ㄉㄜ˙ ㄕˋㄑㄧㄥˊ。
+./custom_character_voice/linghua/processed_103.wav|ㄉㄢˋ ㄖㄨˊㄍㄨㄛˇ ㄕˋ ㄧㄣˉㄨㄟˋ ㄗˋㄐㄧˇ ㄉㄜ˙ ㄕˋㄑㄧㄥˊ, ㄐㄧㄡˋ ㄖㄤˋ ㄓㄡˉㄗㄠˉ ㄉㄜ˙ ㄖㄣˊ ㄉㄢˉㄒㄧㄣˉ ㄉㄜ�� ㄏㄨㄚˋ……
+./custom_character_voice/linghua/processed_104.wav|ㄍㄨㄛˇㄖㄢˊ, ㄨㄛˇ ㄏㄞˊㄕˋ ㄅㄨˋㄋㄥˊ ㄕㄨㄛˉ。
+./custom_character_voice/linghua/processed_105.wav|ㄋㄢˊㄕㄨㄞˋ, ㄓㄣˉㄉㄜ˙ ㄕˋ ㄧˉㄐㄧㄢˋ ㄏㄣˇㄋㄢˊ ㄉㄜ˙ ㄕˋㄑㄧㄥˊ。
+./custom_character_voice/linghua/processed_106.wav|ㄨㄛˇ ㄅㄧˋㄒㄩˉ ㄅㄨˋㄉㄨㄢˋ ㄍㄠˋㄐㄧㄝˋ ㄗˋㄐㄧˇ ㄕˋ ㄕㄣˊㄌㄧˋ ㄐㄧㄚˉ ㄉㄜ˙ ㄉㄚˋ ㄒㄧㄠˇㄐㄧㄝˇ ㄗㄞˋ ㄨˊㄕㄨˋㄕㄨㄤˉ ㄧㄢˇㄐㄧㄥˉ， ㄨˊㄕㄨˋ ㄖㄣˊ ㄉㄜ˙ ㄑㄧˉㄆㄢˋ ㄓˉㄒㄧㄚˋ ㄅㄞˇ ㄔㄨˉ ㄨㄢˊㄇㄟˇㄨˊㄑㄩㄝˉ ㄉㄜ˙ ㄗˉㄊㄞˋ。
+./custom_character_voice/linghua/processed_107.wav|ㄓㄜˋㄧㄤˋ ㄉㄜ˙ ㄨㄛˇ, ㄕˋㄈㄡˇ ㄧㄝˇ ㄍㄞˉ ㄓㄨㄟˉㄑㄧㄡˊ ㄗˋㄐㄧˇ ㄉㄜ˙ ㄩㄢˋㄨㄤˋ ㄋㄜ˙?
+./custom_character_voice/linghua/processed_108.wav|ㄕˋㄈㄡˇ ㄧㄝˇ ㄍㄞˉ… ㄖㄤˋ ㄋㄧˇ ㄌㄧˇㄐㄧㄝˇ ㄨㄛˇ ㄉㄜ˙ ㄒㄧㄣˉㄧˋ ㄋㄜ˙?
+./custom_character_voice/linghua/processed_109.wav|ㄨㄛˇ ㄉㄨㄟˋ ㄧˋㄍㄨㄛˊ ㄌㄧㄠˋㄌㄧˇ ㄏㄣˇ ㄧㄡˇ ㄒㄧㄥˋㄑㄩˋ, ㄙㄨㄟˉㄖㄢˊ ㄏㄣˇㄕㄠˇ ㄧㄡˇ ㄔㄤˊㄕˋ ㄉㄜ˙ ㄐㄧˉㄏㄨㄟˋ。
+./custom_character_voice/linghua/processed_110.wav|ㄖㄨˊㄍㄨㄛˇ ㄕˋ ㄧㄠˋ ㄐㄩˊㄒㄧㄢˋ ㄗㄞˋ ㄉㄠˋㄑㄧˉ ㄌㄧㄠˋㄌㄧˇ ㄋㄟˋ ㄉㄜ˙ ㄏㄨㄚˋ, ㄧㄥˉㄍㄞˉ ㄕˋ ㄔㄚˊ ㄆㄠˋ ㄈㄢˋ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_111.wav|ㄅㄨˋㄋㄥˊ ㄍㄟˇ ㄧˋㄅㄢˉ ㄎㄜˋㄖㄣˊ ㄎㄢˋㄐㄧㄢˋ。
+./custom_character_voice/linghua/processed_112.wav|ㄨㄛˇ ㄧㄝˇ ㄕˋ ㄊㄡˉㄊㄡˉ ㄍㄠˋㄙㄨˋ ㄋㄧˇ ㄉㄜ˙。
+./custom_character_voice/linghua/processed_113.wav|ㄙㄨㄟˉㄖㄢˊ ㄅㄨˊㄕˋ ㄅㄨˋㄋㄥˊ ㄔˉ, ㄉㄢˋ ㄨㄛˇ ㄉㄨㄟˋ ㄉㄨㄥˋㄨˋ ㄓˉㄈㄤˊ ㄏㄨㄛˋ ㄋㄟˋㄗㄤˋ… ㄉㄚˋㄍㄞˋ ㄏㄨㄟˋ ㄧㄡˇㄉㄧㄢˇ…
+./custom_character_voice/linghua/processed_114.wav|ㄑㄧㄥˇ ㄙㄨㄟˊ ㄨㄛˇ ㄌㄞˊ, ㄓˇㄧㄠˋ ㄗㄡˇ ㄧˉㄒㄧㄠˇ ㄉㄨㄢˋㄌㄨˋ, ㄅㄨˊㄏㄨㄟˋ ㄉㄢˉㄍㄜˉ ㄋㄧˇ ㄊㄞˋ ㄓㄤˇㄕˊㄐㄧㄢˉ ㄉㄜ˙。
+./custom_character_voice/linghua/processed_115.wav|ㄉㄚˇㄊㄧㄥˉ ㄉㄠˋ ㄋㄧˇ ㄉㄜ˙ ㄕㄥˉㄖˋ ㄓˉㄏㄡˋ, ㄨㄛˇ ㄐㄧㄡˋ ㄊㄧˊㄑㄧㄢˊ ㄌㄜ˙ ㄧˉㄉㄨㄢˋㄕˊㄐㄧㄢˉ ㄎㄞˉㄕˇ ㄔㄡˊㄅㄟˋ。
+./custom_character_voice/linghua/processed_116.wav|ㄅㄧˇㄑㄧˇ ㄇㄧㄥˊㄍㄨㄟˋ ㄉㄜ˙ ㄌㄧˇㄨˋ ㄨㄛˇ ㄨㄤˋㄗˋ ㄘㄞˉㄘㄜˋ ㄏㄨㄛˋㄒㄩˇ ㄓㄜˋㄧㄤˋ ㄉㄜ˙ ㄌㄧˇㄨˋ ㄏㄨㄟˋ ㄍㄥˋㄏㄜˊ ㄨㄛˇㄇㄣ˙ ㄉㄜ˙ ㄑㄧㄥˊㄧˋ?
+./custom_character_voice/linghua/processed_117.wav|ㄘˇㄘˋ, ㄐㄧㄡˋ ㄑㄧㄥˇ ㄖㄤˋ ㄨㄛˇ ㄧˇㄕㄢˋ ㄨˊㄨㄟˋㄌㄧˇ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_118.wav|ㄕˉㄌㄧˇ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_119.wav|ㄍㄢˇㄒㄧㄝˋ, ㄩˇ ㄋㄧˇ ㄑㄧㄝˉㄘㄨㄛˉ ㄕˇ ㄨㄛˇ ㄕㄡˉㄧˋ ㄌㄧㄤˊㄉㄨㄛˉ, ㄒㄧㄤˉㄒㄧㄣˋ ㄗㄞˋ ㄐㄧㄢˋㄕㄨˋ ㄕㄤˋ ㄧㄝˇ ㄋㄥˊ ㄍㄥˋ ㄐㄧㄣˋ ㄧˉㄅㄨˋ。
+./custom_character_voice/linghua/processed_120.wav|ㄉㄨㄛˉㄎㄨㄟˉ ㄋㄧˇ ㄉㄜ˙ ㄉㄧㄢˇㄅㄛˉ。
+./custom_character_voice/linghua/processed_121.wav|ㄨㄛˇ ㄉㄨㄟˋ ㄗˋㄐㄧˇ ㄉㄜ˙ ㄋㄥˊㄌㄧˋ ㄧㄝˇ ㄌㄧˇㄐㄧㄝˇ ㄉㄜˊ ㄍㄥˋ ㄊㄡˋㄔㄜˋ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_122.wav|ㄍㄥˋㄐㄧㄚˉ ㄧㄡˊㄖㄣˋㄧㄡˇㄩˊ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_123.wav|ㄕㄡˉ ㄈㄥˋㄒㄧㄥˊ ㄕˋㄨˋ ㄓˉㄩˊ, ㄕㄣˋㄓˋ ㄧㄡˇㄎㄨㄥˋ ㄔㄤˊㄕˋ ㄧˉㄒㄧㄝˉ ㄒㄧㄣˉ ㄉㄜ˙ ㄕˋㄑㄧㄥˊ。
+./custom_character_voice/linghua/processed_124.wav|ㄒㄧㄤˇㄧㄠˋ ㄌㄞˊ ㄕˋㄕˋ ㄗˋㄐㄧˇ ㄒㄩㄝˊㄗㄨㄛˋ ㄉㄜ˙ ㄉㄧㄢˇㄒㄧㄣˉ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_125.wav|ㄒㄧㄣˉ ㄧㄡˇㄙㄨㄛˇㄙˉ, ㄙˉㄕˋ ㄖㄨˊㄆㄢˋ。
+./custom_character_voice/linghua/processed_126.wav|ㄊㄨˊㄌㄧㄠˋ ㄈㄨˊㄕˋ ㄕˋ, ㄌㄧㄡˊㄓㄨˋ ㄋㄢˊ。
+./custom_character_voice/linghua/processed_127.wav|ㄅㄠˋㄑㄧㄢˋ, ㄇㄧㄥˊㄇㄧㄥˊ ㄕˋ ㄓˊㄉㄜ˙ ㄍㄠˉㄒㄧㄥˋ ㄉㄜ˙ ㄕˊㄏㄡˋ, ㄨㄛˇ ㄑㄩㄝˋ ㄒㄧㄤˇㄑㄧˇ ㄌㄜ˙ ㄋㄚˋㄇㄜ˙ ㄅㄟˉㄕㄤˉ ㄉㄜ˙ ㄕˊㄎㄜˋ……
+./custom_character_voice/linghua/processed_128.wav|ㄒㄧㄤˉㄔㄨˋ ㄉㄜ˙ ㄕˊㄐㄧㄢˉㄍㄨㄛˋ ㄩˊ ㄔㄤˋㄏㄨㄢˇ ㄐㄧㄥˋㄖㄢˊ ㄖㄤˋ ㄨㄛˇ ㄏㄞˋㄆㄚˋ ㄗㄞˋㄘˋ ㄕˉㄑㄩˋ。
+./custom_character_voice/linghua/processed_129.wav|ㄕˉ… ㄕˉㄊㄞˋ ㄌㄜ˙ ㄧㄚ˙。
+./custom_character_voice/linghua/processed_130.wav|ㄅㄧˇㄨˇ ㄅㄚ˙ ㄕˉㄌㄧˇ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_132.wav|ㄕㄣˊㄌㄧˇㄌㄧㄡˊ  ㄕㄨㄤˉㄇㄧㄝˋ。
+./custom_character_voice/linghua/processed_133.wav|ㄓㄜˋ ㄇㄧˋ ㄑㄧˊ, ㄧㄝˇ ㄙㄨㄢˋ ㄉㄜˊ ㄧˉㄓㄨㄥˇ ㄧㄚˇㄑㄩˋ。
+./custom_character_voice/linghua/processed_134.wav|ㄐㄧㄣˉㄖˋ ㄩㄣˋㄕˋ ㄅㄨˋㄘㄨㄛˋ, ㄨㄛˇㄏㄨㄟˋ ㄓㄣˉㄒㄧˉ ㄓㄜˋ ㄧˉㄈㄣˋ ㄒㄧㄥˋㄩㄣˋ, ㄅㄨˋㄖㄨㄥˊ ㄒㄧㄠˇㄑㄩˋ ㄋㄜ˙。
+./custom_character_voice/linghua/processed_135.wav|ㄨㄛˇ ㄉㄜ˙ ㄉㄨㄟˋㄕㄡˇ, ㄍㄞˉ ㄐㄩㄝˊㄉㄨㄢˋ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_136.wav|ㄕˉㄊㄞˋ ㄌㄜ˙ ㄨㄛˇ ㄏㄞˊㄧㄡˇ ㄨㄟˋㄐㄧㄣˇ ㄓˉㄕˋ ㄖㄤˋ ㄐㄧㄚˉㄗㄨˊ ㄇㄥˊㄒㄧㄡˉ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_137.wav|ㄉㄜ˙ ㄐㄧㄚˉㄏㄨㄛ˙… ㄗㄣˇㄇㄜ˙ ㄏㄨㄟˋ…
+./custom_character_voice/linghua/processed_138.wav|ㄕㄣˊㄌㄧˇㄌㄧㄥˊㄏㄨㄚˊ。
+./custom_character_voice/linghua/processed_140.wav|ㄑㄧㄥˇ ㄘˋㄐㄧㄠˋ。
+./custom_character_voice/linghua/processed_141.wav|ㄑㄧㄥˇ ㄉㄨㄛˉ ㄍㄨㄢˉㄓㄠˋ。

filelists/final_annotation_val.txt

@@ -0,0 +1,137 @@
+./custom_character_voice/linghua/processed_0.wav|ㄉㄠˋㄑㄧˉ ㄕㄣˊㄌㄧˇ ㄌㄧㄡˊㄊㄞˋ ㄉㄠˉㄕㄨˋ ㄐㄧㄝˉ ㄔㄨㄢˊㄕㄣˊ ㄌㄧˇㄌㄧㄥˊㄏㄨㄚˊ ㄘㄢˉㄕㄤˋ。
+./custom_character_voice/linghua/processed_1.wav|ㄑㄧㄥˇ ㄉㄨㄛˉ ㄓˇㄐㄧㄠˋ ㄚ˙。
+./custom_character_voice/linghua/processed_2.wav|ㄓㄜˋㄧㄤˋ ㄧㄡˉㄒㄧㄢˊㄢˉㄨㄣˇ ㄉㄜ˙ ㄕˊㄍㄨㄤˉ, ㄖㄨˊㄍㄨㄛˇ ㄗㄞˋ ㄉㄨㄛˉ ㄧˉㄉㄧㄢˇ ㄐㄧㄡˋ ㄏㄠˇ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_3.wav|ㄨㄛˇ ㄓㄣˉ ㄊㄢˉㄒㄧㄣˉ ㄚ˙。
+./custom_character_voice/linghua/processed_4.wav|ㄐㄧㄡˋ ㄏㄜˊ ㄔㄚˊ ㄧˉㄧㄤˋ, ㄒㄧˋㄒㄧˋㄆㄧㄣˇㄨㄟˋ, ㄘㄞˊㄋㄥˊ ㄌㄧˇㄐㄧㄝˇ ㄑㄧˊㄓㄨㄥˉ ㄈㄥˉㄧㄚˇ。
+./custom_character_voice/linghua/processed_5.wav|ㄉㄡˉ ㄕˋ ㄌㄩˇㄒㄧㄥˊㄓㄜˇ ㄆㄧㄥˊㄖˋ ㄌㄧˇ ㄉㄜ˙ ㄕㄥˉㄏㄨㄛˊ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_6.wav|ㄍㄢˇㄐㄩㄝˊ ㄧㄡˋ ㄉㄨㄛˉ ㄌㄧㄠˇㄐㄧㄝˇ ㄌㄜ˙ ㄋㄧˇ ㄧˉㄒㄧㄝˉ。
+./custom_character_voice/linghua/processed_7.wav|ㄐㄧㄢˋ ㄅㄠˋㄧㄝˋㄇㄧㄥˊ ㄍㄨㄥˉㄏㄨㄞˊ ㄅㄠˇ。
+./custom_character_voice/linghua/processed_8.wav|ㄙㄨㄟˊ ㄨㄛˇ ㄧˉㄊㄨㄥˊ ㄅㄧˋㄩˇ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_9.wav|ㄙㄨㄛˇ ㄉㄚˋㄖㄣˊ, ㄕˋ ㄗㄞˋ ㄙㄨˋㄕㄨㄛˉ ㄕㄣˊㄇㄜ˙ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_10.wav|ㄐㄧˋㄧㄣˊㄓㄨㄤˉㄙㄨˋ, ㄐㄩˊㄍㄠˉㄧㄥˋ ㄑㄩㄥˊㄓˉ。
+./custom_character_voice/linghua/processed_11.wav|ㄣˊ…… ㄇㄟˇㄐㄧㄥˇ ㄉㄤˉㄑㄧㄢˊ, ㄓˇㄔㄚˋ ㄧˉㄏㄨˊ ㄔㄚˊ ㄩˇ ㄓˉ ㄒㄧㄤˉㄔㄣˋ ㄋㄜ˙。
+./custom_character_voice/linghua/processed_12.wav|ㄧㄠˋ ㄑㄩˋ ㄋㄚˇㄅㄧㄢˉ ㄗㄡˇㄗㄡˇ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_13.wav|ㄧㄢˇㄐㄧㄥˉ, ㄒㄧㄤˋ ㄓㄨˋㄈㄥˉ ㄔㄨㄟˉ ㄌㄞˊ ㄉㄜ˙ ㄈㄤˉㄒㄧㄤˋ。
+./custom_character_voice/linghua/processed_14.wav|ㄞˉㄧㄚˉ, ㄏㄣˇ ㄕㄨˉㄈㄨˊ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_15.wav|ㄌㄩˇㄒㄧㄥˊㄓㄜˇ。
+./custom_character_voice/linghua/processed_16.wav|ㄓㄜˋㄧㄤˋ ㄗㄞˋ ㄑㄧㄥˉㄔㄣˊ ㄐㄧㄢˋ ㄋㄧˇ ㄧˊㄇㄧㄢˋ, ㄨㄛˇㄏㄨㄟˋ ㄖㄣˇㄅㄨˊㄓㄨˋ ㄐㄩㄝˊㄉㄜˊ, ㄐㄧㄝˉㄒㄧㄚˋ ㄌㄞˊㄐㄧㄤˉ ㄕˋ ㄕㄨㄣˋㄌㄧˋ ㄉㄜ˙ ㄧˉㄊㄧㄢˉ。
+./custom_character_voice/linghua/processed_18.wav|ㄔㄚˊㄈㄢˋ ㄓˉㄏㄡˋ, ㄋㄢˊㄇㄧㄢˇ ㄌㄩㄝˋㄧㄡˇ ㄎㄨㄣˋㄐㄩㄢˋ ㄕˋㄈㄡˇ ㄧㄡˇ ㄒㄧㄥˋㄓˋ ㄒㄧㄚˋㄆㄢˊ ㄑㄧˊ ㄊㄧˊㄕㄣˊ ㄋㄜ˙?
+./custom_character_voice/linghua/processed_20.wav|ㄏㄨㄟˋ ㄕˋ ㄧˊㄍㄜˋ ㄌㄧㄤˊㄒㄧㄠˉ ㄋㄜ˙。
+./custom_character_voice/linghua/processed_21.wav|ㄓˉㄕˋ ㄇㄥˋ  ㄏㄜˊㄒㄩˉ ㄒㄧㄥˇ。
+./custom_character_voice/linghua/processed_22.wav|ㄅㄨˋㄅㄧˇㄓㄣˉ ㄖㄨˊ, ㄧˉㄒㄧㄤˉㄏㄨㄟˋ。
+./custom_character_voice/linghua/processed_23.wav|ㄉㄠˋㄑㄧˉ ㄇㄨˋㄈㄨˇ ㄕㄜˋ ㄈㄥˋㄒㄧㄥˊ ㄕㄣˊㄌㄧˇㄐㄧㄚˉ, ㄨㄟˋ ㄩˊ ㄉㄠˋㄑㄧˉ ㄇㄧㄥˊㄇㄣˊ ㄓㄨㄥˉ ㄉㄜ˙ ㄅㄧˇㄊㄡˊ ㄓˉ ㄍㄜˊㄨㄟˋ。
+./custom_character_voice/linghua/processed_24.wav|ㄗㄨㄛˋㄨㄟˊ ㄙㄢˉ ㄈㄥˋㄒㄧㄥˊ ㄓˉㄧˉ, ㄓㄤˇㄍㄨㄢˇ ㄐㄧˋㄙˋ ㄏㄨㄛˊㄉㄨㄥˋ ㄩˇ ㄖㄣˊㄨㄣˊ ㄧˋㄕㄨˋ。
+./custom_character_voice/linghua/processed_25.wav|ㄕㄨㄤˉㄑㄧㄣˉ ㄍㄨㄛˋㄕˋ ㄓˉㄏㄡˋ, ㄗㄨˊ ㄋㄟˋ ㄉㄜ˙ ㄉㄚˋㄒㄧㄠˇ ㄕˋㄨˋ ㄅㄧㄢˋ ㄧㄡˊ ㄒㄩㄥˉㄓㄤˇ ㄏㄜˊ ㄨㄛˇ ㄔㄥˊㄉㄢˉ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_26.wav|ㄏㄣˇㄉㄨㄛˉ ㄖㄣˊㄧㄣˉ ㄨㄟˋ ㄨㄛˇ ㄕˋ ㄅㄞˊㄌㄨˋ ㄍㄨㄥˉㄓㄨˇ, ㄕˋ ㄕㄜˋ ㄈㄥˋㄒㄧㄥˊ ㄕㄣˊㄌㄧˇㄐㄧㄚˉ ㄉㄜ˙ ㄉㄚˋ ㄒㄧㄠˇㄐㄧㄝˇ, ㄦˊ ㄐㄧㄥˋㄓㄨㄥˋ ㄨㄛˇ。
+./custom_character_voice/linghua/processed_27.wav|ㄊㄚˉㄇㄣ˙ ㄙㄨㄛˇ ㄐㄧㄥˋㄓㄨㄥˋ ㄉㄜ˙, ㄓˇㄕˋ ㄨㄛˇ ㄙㄨㄛˇ ㄕㄣˉㄔㄨˋ ㄉㄜ˙ ㄉㄧˋㄨㄟˋ, ㄩˇㄌㄧㄥˊㄏㄨㄚˊ ㄨㄛˇ ㄕˋ ㄗㄣˇㄧㄤˋ ㄉㄜ˙ ㄖㄣˊ ㄅㄧㄥˋ ㄨˊ ㄍㄨㄢˉㄒㄧˋ。
+./custom_character_voice/linghua/processed_28.wav|ㄨㄛˇ ㄒㄧㄤˇ, ㄋㄥˊ ㄓㄣˉㄓㄥˋ ㄗㄡˇㄐㄧㄣˋ ㄨㄛˇ ㄉㄜ˙, ㄏㄨㄛˋㄒㄩˇ ㄓˇㄧㄡˇ, ㄖㄨˊㄐㄧㄣˉ ㄉㄜ˙ ㄨㄛˇ。
+./custom_character_voice/linghua/processed_29.wav|ㄧˉㄐㄧㄡˋ ㄒㄧㄤˇ ㄔㄥˊㄨㄟˋ ㄓˊㄉㄜ˙ ㄉㄚˋㄐㄧㄚˉ ㄒㄧㄣˋㄖㄣˋ ㄉㄜ˙ ㄖㄣˊ。
+./custom_character_voice/linghua/processed_30.wav|ㄍㄨˇㄨˇ ㄨㄛˇ ㄉㄜ˙ ㄩㄢˊㄧㄣˉ, ㄧˇ ㄅㄨˋㄗㄞˋ ㄕˋ ㄐㄧㄢˉㄕㄤˋ ㄉㄜ˙ ㄗㄜˊㄖㄣˋ, ㄏㄨㄛˋ ㄊㄚˉㄖㄣˊ ㄉㄜ˙ ㄑㄧˉㄉㄞˋ。
+./custom_character_voice/linghua/processed_31.wav|ㄕˋ ㄧㄣˉㄨㄟˋ…… ㄋㄧˇ ㄧㄝˇ ㄕˋ ㄓㄜˋㄧㄤˋ ㄉㄜ˙ ㄖㄣˊ ㄚ˙。
+./custom_character_voice/linghua/processed_32.wav|ㄖㄨˊㄍㄨㄛˇ ㄋㄧㄣˊ ㄧㄡˇㄎㄨㄥˋ, ㄨㄛˇㄇㄣ˙ ㄧˉㄅㄨˋ ㄇㄨˋㄌㄨˋ ㄔㄚˊㄕˋ ㄖㄨˊㄏㄜˊ?
+./custom_character_voice/linghua/processed_33.wav|ㄗㄞˋ ㄓㄜˋㄧㄤˋ ㄊㄧㄢˊㄐㄧㄥˋ ㄉㄜ˙ ㄖˋㄗ˙, ㄌㄩㄝˋㄐㄧㄚˉ ㄐㄧㄠˉㄌㄧㄡˊ ㄔㄚˊㄧˋ ㄒㄧㄣˉㄉㄜˊ, ㄒㄧㄤˇㄌㄞˊ ㄕˋ ㄆㄛˇㄐㄩˋ ㄧㄚˇㄑㄩˋ ㄉㄜ˙。
+./custom_character_voice/linghua/processed_34.wav|ㄖㄨˊㄍㄨㄛˇ ㄧㄡˇ ㄐㄧˉㄏㄨㄟˋ ㄉㄜ˙ㄏㄨㄚˋ, ㄨㄛˇㄒㄧㄤˇㄕˋ ㄓㄜ˙ ㄏㄜˊ ㄋㄧˇ ㄍㄨㄥˋㄉㄨˋ ㄧˋㄍㄨㄛˊ ㄉㄜ˙ ㄐㄧㄝˊㄖˋ。
+./custom_character_voice/linghua/processed_35.wav|ㄗㄨㄣˉㄒㄩㄣˊ ㄉㄜ˙ ㄉㄤˉㄉㄧˋ ㄈㄥˉㄙㄨˊ， ㄌㄧˇㄧˊ ㄍㄨㄟˉㄈㄢˋ, ㄏㄞˊㄧㄡˇ ㄅㄢˋㄕㄡˇ ㄌㄧˇ ㄉㄜ˙ ㄊㄨㄟˉㄐㄧㄢˋ。
+./custom_character_voice/linghua/processed_36.wav|ㄎㄜˇㄧˇ ㄇㄚˊㄈㄢˊ ㄋㄧˇ… ㄧˉㄧˉ ㄓˇㄉㄠˇ ㄨㄛˇ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_37.wav|ㄕㄣˊㄓˉㄧㄢˇ, ㄐㄧˊ ㄕˋ ㄒㄩㄥˉㄏㄨㄞˊㄉㄚˋㄓˋ ㄓˉㄖㄣˊ ㄙㄨㄛˇ ㄏㄨㄛˋ ㄉㄜ˙ ㄧㄥˉㄕㄡˋ。
+./custom_character_voice/linghua/processed_38.wav|ㄖㄨˊㄍㄨㄛˇ ㄨㄣˋ ㄨㄛˇ ㄧㄡˇ ㄕㄣˊㄇㄜ˙ ㄓˋㄒㄧㄤˋ ㄉㄜ˙ㄏㄨㄚˋ。
+./custom_character_voice/linghua/processed_39.wav|ㄓㄜˋㄍㄜˋ ㄏㄞˊㄕˋ ㄅㄠˇㄇㄧˋ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_40.wav|ㄓˇㄕˋ ㄧˊㄍㄜˋ ㄨㄟˉㄅㄨˋㄗㄨˊㄉㄠˋ ㄉㄜ˙ ㄇㄥˋㄒㄧㄤˇ ㄅㄚˋㄌㄜ˙。
+./custom_character_voice/linghua/processed_41.wav|ㄔㄤˊㄕㄨㄛˉ  ㄔㄢˊ ㄔㄚˊ ㄧˉㄨㄟˋ。
+./custom_character_voice/linghua/processed_42.wav|ㄐㄧㄢˋ ㄔㄢˊ ㄧˋ ㄖㄨˊ。
+./custom_character_voice/linghua/processed_43.wav|ㄋㄚˋㄇㄜ˙ ㄐㄧㄢˋ ㄏㄜˊ ㄔㄚˊ, ㄧㄡˋ ㄕˋ ㄕㄣˊㄇㄜ˙ ㄍㄨㄢˉㄒㄧˋ ㄋㄜ˙?
+./custom_character_voice/linghua/processed_44.wav|ㄋㄧˇ ㄗㄞˋ ㄔㄥˊㄓㄨㄥˉ, ㄐㄧㄢˋㄍㄨㄛˋ ㄎㄨˉㄨㄟˇ ㄉㄜ˙ ㄧㄥˉㄏㄨㄚˉㄕㄨˋ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_45.wav|ㄎㄨˉㄓˉ ㄇㄟˇ  ㄖㄤˋ ㄨㄛˇ ㄒㄧㄤˇㄉㄠˋ ㄔㄨㄣˉㄊㄧㄢˉ ㄕㄥˋㄎㄞˉ ㄓˉㄐㄧㄥˇ。
+./custom_character_voice/linghua/processed_46.wav|ㄅㄨˋㄍㄨㄛˋ, ㄅㄧㄝˊㄖㄣˊ ㄙˋㄏㄨˉ ㄅㄧㄥˋ ㄅㄨˋ ㄓㄜˋㄇㄜ˙ ㄒㄧㄤˇ。
+./custom_character_voice/linghua/processed_47.wav|ㄎㄞˉㄏㄨㄚˉ ㄉㄜ˙ ㄧㄣˉㄕㄨˋㄏㄨㄟˋ ㄅㄟˋ ㄧˊㄗㄡˇ。
+./custom_character_voice/linghua/processed_48.wav|ㄐㄧㄡˋㄙㄨㄢˋ ㄧˉㄘˋ ㄧㄝˇㄏㄠˇ, ㄓㄣˉㄒㄧㄤˇ ㄎㄢˋㄉㄠˋ ㄊㄚˉ ㄗㄞˋㄘˋ ㄎㄞˉㄈㄤˋ。
+./custom_character_voice/linghua/processed_49.wav|ㄕˋ ㄨㄛˇ ㄏㄣˇ ㄓㄨㄥˋㄧㄠˋ ㄉㄜ˙ ㄆㄥˊㄧㄡˇ。
+./custom_character_voice/linghua/processed_50.wav|ㄊㄧㄢˉㄌㄥˇ ㄏㄜˊ ㄧㄤˊㄍㄨㄤˉ, ㄗㄨㄥˇㄕˋ ㄍㄢˇㄖㄢˇ ㄓㄜ˙ ㄨㄛˇ。
+./custom_character_voice/linghua/processed_51.wav|ㄧˋㄧˋ ㄕㄤˋ ㄌㄞˊㄕㄨㄛˉ, ㄊㄚˉ ㄐㄧㄡˋ ㄒㄧㄤˋㄕˋ ㄨㄛˇ ㄉㄜ˙ ㄌㄧㄥˋ ㄧˊㄍㄜˋ ㄒㄩㄥˉㄓㄤˇ ㄧˉㄧㄤˋ。
+./custom_character_voice/linghua/processed_52.wav|ㄔㄥˊㄨㄟˋ ㄌㄜ˙ ㄕㄣˊㄌㄧˋ ㄐㄧㄚˉ ㄉㄜ˙ ㄧˉㄩㄢˊ。
+./custom_character_voice/linghua/processed_53.wav|ㄌㄧㄥˇ ㄈㄥˋㄒㄧㄥˊ ㄍㄨㄥˉㄗㄨㄛˋ ㄉㄜ˙ ㄐㄧㄡˇㄊㄧㄠˊ ㄕㄚˉㄌㄨㄛˊ, ㄊㄚˉ ㄗㄨㄥˇㄕˋ ㄧˉㄌㄧㄢˇ ㄧㄢˊㄙㄨˋ。
+./custom_character_voice/linghua/processed_54.wav|ㄊㄚˉ ㄘㄥˊㄐㄧㄥˉ ㄑㄧˇㄍㄨㄛˋ ㄐㄧˇㄘˋ ㄓㄥˉㄉㄨㄢˉ。
+./custom_character_voice/linghua/processed_55.wav|ㄊㄚˉ ㄅㄣˇㄓˋ ㄕˋ ㄓㄨㄥˉㄧˋ ㄓˉㄕˋ, ㄓㄜˋㄉㄧㄢˇ ㄨˊㄎㄜˇㄈㄡˇㄖㄣˋ。
+./custom_character_voice/linghua/processed_56.wav|ㄙㄨˉㄧㄝˇ ㄩㄝˋㄒㄧㄢˉㄕㄥˉ。
+./custom_character_voice/linghua/processed_57.wav|ㄊㄚˉ ㄏㄜˊ ㄐㄧㄡˇㄊㄧㄠˊ ㄒㄧㄠˇㄐㄧㄝˇ ㄧˉㄧㄤˋ, ㄕˋ ㄊㄧㄢˉㄌㄧㄥˇ ㄈㄥˋㄒㄧㄥˊ ㄉㄜ˙ ㄖㄣˊ。
+./custom_character_voice/linghua/processed_58.wav|ㄊㄚˉ…… ㄧㄝˇ ㄕˋ ㄧˊㄍㄜˋ ㄏㄣˇ ㄧㄡˇ ㄩㄢˊㄗㄜˊ ㄉㄜ˙ ㄖㄣˊ。
+./custom_character_voice/linghua/processed_59.wav|ㄓㄜˋㄒㄧㄝˉ ㄩㄢˊㄗㄜˊ ㄉㄜ˙ ㄐㄧㄢˉㄔˊ, ㄕㄣˋㄓˋ ㄅㄧˇ ㄐㄧㄡˇㄊㄧㄠˊ ㄒㄧㄠˇㄐㄧㄝˇ ㄍㄥˋ ㄓˊㄓㄨㄛˊ。
+./custom_character_voice/linghua/processed_60.wav|ㄅㄨˋㄍㄨㄛˋ, ㄕㄣˊㄇㄜ˙ ㄕˋ ㄧㄥˉㄍㄞˉ ㄅㄟˋ ㄙㄨㄢˋ ㄗㄞˋ ㄓㄜˋㄒㄧㄝˉ ㄩㄢˊㄗㄜˊ ㄓˉㄋㄟˋ…… ㄨㄛˇ ㄒㄧㄤˇ, ㄏㄨㄛˋㄒㄩˇ ㄓˇㄧㄡˇ ㄌㄨˋ ㄧㄝˇ ㄩㄢ�� ㄒㄧㄢˉㄕㄥˉ ㄗˋㄐㄧˇ ㄓˉㄉㄠˋ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_61.wav|ㄒㄧㄠˇ ㄧㄡˋㄓˋ ㄏㄞˊㄗ˙, ㄗㄨㄟˋㄐㄧㄣˋ ㄧㄡˇ ㄇㄟˊ ㄧㄡˇㄍㄟˇ ㄋㄧˇ ㄊㄧㄢˉㄕㄣˊㄇㄜ˙ ㄇㄚˊㄈㄢˊ ㄋㄜ˙?
+./custom_character_voice/linghua/processed_62.wav|ㄖㄨˊㄍㄨㄛˇ ㄎㄢˋㄐㄧㄢˋ ㄊㄚˉ ㄊㄡˉ ㄌㄢˇ, ㄎㄜˇㄧˇ ㄓˊㄐㄧㄝˉ ㄍㄠˋㄙㄨˋ ㄨㄛˇ。
+./custom_character_voice/linghua/processed_63.wav|ㄌㄧˇㄙㄨㄛˇ ㄉㄤˉㄖㄢˊ ㄉㄜ˙ ㄎㄢˋㄈㄚˇ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_64.wav|ㄅㄨˋㄍㄞˉ ㄧㄡˊ ㄨㄛˇ ㄉㄥˇ ㄒㄧㄚˋㄕㄨˇ ㄙㄨㄟˊㄧˋ ㄧˋㄌㄨㄣˋ。
+./custom_character_voice/linghua/processed_65.wav|ㄏㄥˉ, ㄐㄧㄤˉㄐㄩㄣˉ ㄉㄚˋㄖㄣˊ ㄊㄚˉ ㄗㄞˋ ㄓㄨㄟˉㄑㄧㄡˊ ㄩㄥˇㄏㄥˊ ㄓˉ ㄌㄨˋㄕㄤˋ, ㄎㄜˇㄋㄥˊ ㄧㄝˇ ㄏㄣˇ ㄍㄨˉㄉㄨˊ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_66.wav|ㄧˉㄉㄠˉ, ㄅㄧㄥˋㄑㄧㄝˇ ㄏㄨㄛˊ ㄌㄜ˙ ㄒㄧㄚˋㄌㄞˊ。 ㄍㄞˉ ㄕㄨㄛˉ ㄕˋ ㄎㄢˉㄔㄥˉ ㄨㄟˇㄧㄝˋ ㄉㄜ˙ ㄐㄧㄥˉㄌㄧˋ ㄌㄜ˙ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_67.wav|ㄙㄨㄟˉㄖㄢˊ ㄉㄨㄟˋ ㄨㄛˇ ㄌㄞˊㄕㄨㄛˉ, ㄊㄚˉ ㄕˋ ㄓㄣˉㄓㄥˋ ㄉㄜ˙ ㄕㄣˊㄇㄧㄥˊ。
+./custom_character_voice/linghua/processed_68.wav|ㄎㄜˇㄧˇ ㄍㄥˉㄍㄞˇ ㄉㄠˋㄑㄧˉ ㄉㄜ˙ ㄇㄧㄥˋㄩㄣˋ。
+./custom_character_voice/linghua/processed_69.wav|ㄉㄢˋㄕˋ ㄖㄨˊㄍㄨㄛˇ ㄕˋ ㄏㄜˊ ㄋㄧˇ ㄑㄧˇ ㄌㄜ˙ ㄔㄨㄥˉㄊㄨˉ ㄉㄜ˙ㄏㄨㄚˋ, ㄨㄛˇ ㄧˊㄉㄧㄥˋ ㄏㄨㄟˋ ㄓㄢˋ ㄗㄞˋ ㄋㄧˇ ㄓㄜˋ ㄧˉㄅㄧㄢˉ ㄉㄜ˙。
+./custom_character_voice/linghua/processed_70.wav|ㄅㄚˉㄓㄨㄥˋ ㄍㄨㄥˉㄙˉ ㄉㄚˋㄖㄣˊ ㄉㄜ˙ ㄏㄜˊㄗㄨㄛˋ ㄒㄧㄤˋㄌㄞˊ ㄏㄣˇ ㄩˊㄎㄨㄞˋ。
+./custom_character_voice/linghua/processed_71.wav|ㄎㄢˋ, ㄘㄠˉㄅㄢˋ ㄐㄧㄝˊㄑㄧㄥˋㄑㄧㄥˋㄉㄧㄢˇ ㄈㄟˉㄔㄤˊ ㄌㄠˊㄕㄣˊㄈㄟˋㄌㄧˋ, ㄧㄥˊㄕㄡˉ ㄉㄨㄛˉㄅㄢˋ ㄧㄝˇ ㄅㄨˋ ㄏㄠˇㄎㄢˋ。
+./custom_character_voice/linghua/processed_72.wav|ㄅㄚˉㄓㄨㄥˋ ㄍㄨㄥˉㄙˉ ㄉㄚˋㄖㄣˊ ㄘㄠˉㄅㄢˋ ㄉㄜ˙ ㄔㄢˇㄧㄝˋ, ㄓㄣˉㄉㄜ˙ ㄐㄧˋ ㄈㄥˉㄧㄚˇ ㄧㄡˋ ㄧㄡˇ ㄕㄡˉㄔㄥˊ。
+./custom_character_voice/linghua/processed_73.wav|ㄗㄨㄛˋㄨㄟˋ ㄕㄣˊㄌㄧˋ ㄐㄧㄚˉ ㄉㄜ˙ ㄐㄧㄚˉㄓㄨˇ, ㄒㄩㄥˉㄓㄤˇ ㄆㄧㄥˊㄖˋ ㄒㄩˉㄧㄠˋ ㄔㄨˉㄇㄧㄢˋ ㄓㄨˉㄉㄨㄛˉㄔㄤˇ ㄏㄜˊ。
+./custom_character_voice/linghua/processed_74.wav|ㄨㄛˇ ㄙㄨㄟˉ ㄐㄧㄣˇㄌㄧˋ ㄈㄣˉㄉㄢˉ ㄒㄩㄥˉㄓㄤˇ ㄐㄧㄢˉㄕㄤˋ ㄙㄨㄛˇ ㄈㄨˋㄉㄢˉ ㄉㄜ˙ ㄗㄜˊㄖㄣˋ, ㄑㄩㄝˋ ㄧㄝˇ ㄨˊㄈㄚˇㄏㄨㄢˇ ㄏㄜˊ ㄊㄚˉ ㄔㄤˊㄋㄧㄢˊ ㄐㄧˉㄧㄚˉ ㄗㄞˋ ㄕㄣˉ ㄉㄜ˙ ㄆㄧˊㄐㄩㄢˋ。
+./custom_character_voice/linghua/processed_75.wav|ㄨㄛˇ ㄧㄝˇ ㄕˋ ㄉㄤˉㄕˊ ㄨㄟˋㄌㄜ˙ ㄇㄟˋㄇㄟˋ ㄓㄨㄛˊㄒㄧㄤˇ, ㄒㄧˉㄨㄤˋ ㄋㄧˇ ㄋㄥˊ ㄑㄩㄢˋㄧㄢˊ, ㄖㄤˋ ㄒㄩㄥˉㄓㄤˇ ㄉㄨㄛˉㄉㄨㄛˉ ㄓㄨˋㄧˋ ㄕㄣˉㄊㄧˇ ㄚ˙。
+./custom_character_voice/linghua/processed_76.wav|ㄐㄧㄝˊㄑㄧㄥˋㄑㄧㄥˋㄉㄧㄢˇ ㄕˋ ㄕㄥˋ ㄈㄥˋㄒㄧㄥˊ ㄏㄜˊ ㄨˉㄋㄩˇ ㄓㄨㄥˉ ㄉㄜ˙ ㄗㄜˊㄖㄣˋ。
+./custom_character_voice/linghua/processed_77.wav|ㄗㄨㄛˋㄨㄟˋ ㄧㄢˉㄏㄨㄛˇ ㄓㄨㄢˉㄐㄧㄚˉ, ㄧㄝˇ ㄉㄜ˙ ㄑㄩㄝˋㄋㄥˊ ㄖㄤˋ ㄑㄧˋㄈㄣˉ ㄖㄜˋㄌㄧㄝˋ ㄑㄧˇㄌㄞˊ。
+./custom_character_voice/linghua/processed_78.wav|ㄏㄜˊㄗㄨㄛˋ ㄉㄨㄛˉ ㄌㄜ˙, ㄧㄣˉㄦˊ ㄐㄧㄢˋㄐㄧㄢˋ ㄕㄨˊㄌㄨㄛˋ。
+./custom_character_voice/linghua/processed_79.wav|ㄓˋㄢˉ ㄏㄜˊ ㄒㄧㄠˉㄈㄤˊ ㄨㄣˋㄊㄧˊ ㄇㄚ˙? ㄨㄛˇㄇㄣ˙ ㄧㄝˇ ㄏㄨㄟˋ ㄧˉ ㄅㄧㄥˋ ㄋㄚˋㄖㄨˋ ㄩˋㄒㄧㄢˉ ㄍㄨㄟˉㄏㄨㄚˋ ㄉㄜ˙。
+./custom_character_voice/linghua/processed_80.wav|ㄓㄜˋㄒㄧㄝˉ ㄩㄢˊㄧㄣˉ ㄦˊ ㄎㄢˋㄅㄨˊㄉㄠˋ ㄧㄥˉㄏㄨㄚˉ, ㄘㄞˊ ㄏㄨㄟˋ ㄖㄤˋ ㄖㄣˊㄇㄣ˙ ㄒㄧㄣˉㄓㄨㄥˉ ㄌㄧㄡˊㄒㄧㄚˋ ㄑㄩㄝˉㄏㄢˋ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_81.wav|ㄋㄧˇ ㄉㄜ˙ ㄑㄧㄥˇㄑㄧㄡˊ, ㄉㄨㄟˋ ㄨㄛˇ ㄌㄞˊㄕㄨㄛˉ ㄏㄣˇ ㄊㄜˋㄅㄧㄝˊ ㄋㄜ˙ ㄐㄧˋㄖㄢˊ ㄅㄚˇ ㄋㄧˇ ㄉㄤˋㄗㄨㄛˋ ㄆㄥˊㄧㄡˇ, ㄨㄛˇ ㄧㄝˇ ㄧㄥˉ ㄊㄢˇㄔㄥˊㄧˇㄉㄞˋ。
+./custom_character_voice/linghua/processed_82.wav|ㄅㄨˋㄍㄨㄛˋ, ㄕˋㄍㄨㄢˉ ㄕㄣˊㄌㄧˇㄐㄧㄚˉ ㄉㄜ˙ ㄇㄧˋㄇㄧˋ, ㄏㄞˊ ㄒㄧˉㄨㄤˋ ㄋㄧˇ ㄋㄥˊ ㄕㄡˇㄎㄡˇㄖㄨˊㄆㄧㄥˊ。
+./custom_character_voice/linghua/processed_83.wav|ㄕㄣˊㄌㄧˋ ㄐㄧㄚˉㄧㄣˉ ㄨㄟˋ ㄇㄟˊㄋㄥˊ ㄅㄠˇㄏㄨˋ ㄏㄠˇ ㄍㄨㄛˊㄅ���ˇㄐㄧˊ ㄅㄧㄝˊ ㄉㄜ˙ ㄉㄠˉㄍㄨㄥˉ, ㄗㄠˉㄕㄡˋ ㄌㄜ˙ ㄅㄨˋㄒㄧㄠˇ ㄉㄜ˙ ㄔㄨㄥˉㄐㄧˉ。
+./custom_character_voice/linghua/processed_84.wav|ㄅㄧㄝˊㄖㄣˊ ㄓㄨㄥˉ ㄧㄣˉㄇㄡˊ ㄙㄨㄢˋㄐㄧˋ ㄨㄛˇㄇㄣ˙ ㄕㄜˊㄙㄨㄣˇ ㄌㄜ˙ ㄓㄨˉㄉㄨㄛˉ ㄔㄣˊ ㄒㄧㄚˋ ㄕㄡˋㄉㄠˋ ㄒㄩˇㄉㄨㄛˉ ㄗㄜˊㄈㄚˊ。
+./custom_character_voice/linghua/processed_85.wav|ㄕㄣˋㄓˋ ㄧㄣˉ ㄓˉ ㄗㄠˇㄕㄨㄞˉ ㄕㄣˊㄌㄧˋ ㄐㄧㄚˉ ㄗㄞˋ ㄇㄨˋㄈㄨˇ ㄓㄨㄥˉ ㄉㄜ˙ ㄉㄧˋㄨㄟˋ ㄧㄝˇ ㄧˉㄌㄨㄛˋㄑㄧㄢˉㄓㄤˋ。
+./custom_character_voice/linghua/processed_86.wav|ㄏㄠˇ ㄗㄞˋ ㄒㄩㄥˉㄓㄤˇ ㄐㄧˋㄖㄣˋㄏㄡˋㄌㄧˋ ㄨㄢˇ ㄎㄨㄤˊㄌㄢˊ。
+./custom_character_voice/linghua/processed_87.wav|ㄐㄧㄚˉㄇㄣˊ ㄙㄨㄟˉ ㄧˇ ㄈㄨˋ ㄒㄧㄥˋ, ㄕㄜˋㄈㄥˉㄒㄧㄥˇ ㄧˉㄒㄧˉ ㄧㄝˇ ㄕㄤˋㄒㄧㄚˋ ㄑㄧˊㄒㄧㄣˉ, ㄉㄢˋ ㄉㄚˋㄕˋ ㄧㄠˋㄕˋ ㄈㄤˉㄇㄧㄢˋ ㄖㄥˊㄎㄠˋ ㄒㄩㄥˉㄓㄤˇ ㄉㄧㄥˋㄉㄨㄛˊ。
+./custom_character_voice/linghua/processed_88.wav|ㄊㄚˉㄇㄣ˙ ㄙㄨㄛˇㄔㄨㄢˊ ㄉㄜ˙ ㄉㄨㄢˋ ㄉㄠˉ ㄓˉㄕㄨˋ, ㄧㄝˇ ㄧㄣˉ ㄒㄧㄥˉㄒㄧㄤˋ， ㄩㄥˋㄊㄨˊ， ㄎㄨㄤˋㄓˊ， ㄌㄨˊㄏㄨㄛˇ ㄏㄨㄢˊㄐㄧㄥˋ， ㄖㄣˊ ㄓˉ ㄒㄧㄥˋㄍㄜˊ， ㄩㄢˊㄙㄨˋ ㄅㄧㄢˋㄏㄨㄚˋ ㄉㄜ˙ ㄅㄨˋㄊㄨㄥˊ ㄦˊ ㄧㄡˇㄙㄨㄛˇ ㄑㄩˉㄈㄣˉ。
+./custom_character_voice/linghua/processed_89.wav|ㄕˋ ㄉㄠˉ ㄍㄨㄥˉ ㄓˉㄐㄧㄢˉ ㄙㄨㄛˇㄕㄨㄛˉ ㄉㄜ˙ ㄌㄟˊㄉㄧㄢˋ ㄨˇㄔㄨㄢˉ。
+./custom_character_voice/linghua/processed_90.wav|ㄘㄤˊㄇㄧㄥˊ ㄉㄠˉ ㄉㄜ˙ ㄉㄠˉㄍㄨㄥˉ, ㄧㄝˇ ㄅㄟˋ ㄙㄨㄢˋㄗㄨㄛˋ ㄕˋ ㄅㄣˇㄌㄧㄥˇ ㄊㄨㄥˉㄕㄣˊ ㄉㄜ˙ ㄕㄣˊㄕˋ ㄒㄧㄤˉㄍㄨㄢˉ ㄖㄣˊㄩㄢˊ, ㄍㄨㄟˉㄕㄨˇ ㄊㄨㄥˇㄔㄡˊ ㄨㄣˊㄏㄨㄚˋ， ㄧˋㄕㄨˋ， ㄐㄧˋㄙˋ ㄉㄜ˙ ㄕㄜˋㄈㄥˋ ㄒㄧㄥˊㄧˋ ㄆㄞˋ ㄍㄨㄢˇㄌㄧˇ。
+./custom_character_voice/linghua/processed_91.wav|ㄔㄨˉㄒㄧㄢˋ ㄌㄜ˙ ㄉㄠˉㄍㄨㄥˉ ㄅㄟˋㄆㄢˋ ㄉㄜ˙ ㄕˋㄑㄧㄥˊ, ㄗˋㄖㄢˊ ㄐㄧㄡˋㄕˋ ㄕㄣˊㄌㄧˇㄐㄧㄚˉ ㄉㄜ˙ ㄉㄨˊㄅㄢˋ ㄅㄨˋㄌㄧˋ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_92.wav|ㄉㄨㄟˋ ㄨㄛˇ ㄌㄞˊ ㄕㄨㄛˉ, ㄇㄨˇㄑㄧㄣˉ ㄕˋ ㄧˋㄧˋ ㄈㄟˉㄈㄢˊ ㄉㄜ˙ ㄘㄨㄣˊㄗㄞˋ。
+./custom_character_voice/linghua/processed_93.wav|ㄈㄨˊㄓㄨㄤˉ, ㄧㄡˉㄧㄚˇ, ㄨˊㄌㄨㄣˋ ㄩˋㄉㄠˋ ㄗㄣˇㄧㄤˋ ㄉㄜ˙ ㄐㄩˊㄇㄧㄢˋ, ㄉㄡˉ ㄏㄨㄟˋ ㄌㄨˋㄔㄨˉ ㄔㄣˊㄐㄧㄣˋ ㄉㄜ˙ ㄒㄧㄠˋㄖㄨㄥˊ, ㄧˇ ㄘㄨㄥˊㄖㄨㄥˊㄅㄨˋㄆㄛˋ ㄉㄜ˙ ㄊㄞˋㄉㄨˋ, ㄘㄠˉㄔˊ ㄓㄜ˙ ㄕㄣˇㄌㄧˇ ㄐㄧㄚˉ ㄉㄚˋㄉㄚˋㄒㄧㄠˇㄒㄧㄠˇ ㄉㄜ˙ ㄕˋㄨˋ。
+./custom_character_voice/linghua/processed_94.wav|ㄍㄢˇㄑㄧㄥˊ ㄕˋ ㄨㄢˊㄇㄟˇ ㄉㄜ˙ ㄏㄨㄚˋㄕㄣˉ ㄧㄝˇ ㄅㄨˋ ㄨㄟˋㄍㄨㄛˋ。
+./custom_character_voice/linghua/processed_95.wav|ㄉㄢˋ ㄗˋㄘㄨㄥˊ ㄊㄚˉ ㄌㄧˊㄕˋ ㄉㄜ˙ ㄋㄚˋ ㄧˉㄎㄜˋㄑㄧˇ, ㄨㄛˇ ㄐㄧㄡˋ ㄕㄣˉㄑㄧㄝˋ ㄉㄧˋ ㄧˋㄕˊ ㄉㄠˋ, ㄨㄛˇ ㄧˇㄐㄧㄥˉ ㄅㄨˊㄕˋ ㄋㄚˋㄍㄜˋ ㄎㄜˇㄧˇ ㄉㄨㄛˇ ㄗㄞˋ ㄇㄨˇㄑㄧㄣˉ ㄕㄣˉㄏㄡˋ ㄉㄜ˙ ㄒㄧㄠˇㄌㄧㄥˊㄏㄨㄚˉ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_96.wav|ㄩㄢˊㄌㄞˊ ㄧㄠˋㄕㄨㄛˉ ㄉㄜ˙ㄏㄨㄚˋ, ㄎㄜˇㄋㄥˊ ㄅㄨˋㄊㄞˋ ㄈㄨˊㄏㄜˊ ㄉㄠˋㄑㄧㄝˋ ㄇㄨˋㄈㄨˇ ㄕㄜˋ ㄈㄥˉㄒㄧㄥˊ ㄕㄣˊㄌㄧˋ ㄐㄧㄚˉ ㄉㄜ˙ ㄕㄣˉㄈㄣˋ。
+./custom_character_voice/linghua/processed_97.wav|ㄅㄨˋㄍㄨㄛˋ, ㄐㄧㄡˋ ㄨㄛˇ ㄉㄜ˙ ㄆㄢˋㄉㄨㄢˋ, ㄋㄧˇ ㄧㄥˉㄍㄞˉ ㄋㄥˊ ㄖㄨㄥˊㄒㄩˇ ㄨㄛˇ ㄒㄧㄠˇㄒㄧㄠˇㄉㄜ˙ ㄖㄣˋㄒㄧㄥˋ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_98.wav|ㄕㄠˉㄨㄟˉ ㄧㄡˇㄉㄧㄢˇㄌㄟˋ ㄌㄜ˙ ㄎㄜˇㄧˇ ㄖㄤˋ ㄨㄛˇ ㄎㄠˋ ㄧˉㄒㄧㄚˋ ㄋㄧˇ ㄉㄜ˙ ㄐㄧㄢˉㄅㄤˇ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_99.wav|ㄐㄧㄡˋㄐㄧㄡˋ。
+./custom_character_voice/linghua/processed_100.wav|ㄨㄛˇ ㄉㄡˉ ㄏㄣˇ ㄒㄧˇㄏㄨㄢˉ ㄧㄚˇㄩㄝˋ ㄕˉㄘˊ ㄑㄧˊㄧˋ ㄩˇ ㄨˇㄉㄠˇ ㄋㄧˇ ㄧㄝˇ ㄍㄢˇㄒㄧㄥˋㄑㄩˋ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_102.wav|…… ㄕˋ ㄚ˙, ㄓˇㄧㄠˋ ㄕㄥˉㄏㄨㄛˊ ㄗㄞˋ ㄓㄜˋㄍㄜˋ ㄕˋㄐㄧㄝˋ ㄕㄤˋ, ㄐㄧㄡˋ ㄋㄢˊㄇㄧㄢˇ ㄩˋㄉㄠˋ ㄓㄨㄥˇㄓㄨㄥˇ ㄅㄨˋㄖㄨˊㄧˋ ㄉㄜ˙ ㄕˋㄑㄧㄥˊ。
+./custom_character_voice/linghua/processed_103.wav|ㄉㄢˋ ㄖㄨˊㄍㄨㄛˇ ㄕˋ ㄧㄣˉㄨㄟˋ ㄗˋㄐㄧˇ ㄉㄜ˙ ㄕˋㄑㄧㄥˊ, ㄐㄧㄡˋ ㄖㄤˋ ㄓㄡˉㄗㄠˉ ㄉㄜ˙ ㄖㄣˊ ㄉㄢˉㄒㄧㄣˉ ㄉㄜ�� ㄏㄨㄚˋ……
+./custom_character_voice/linghua/processed_104.wav|ㄍㄨㄛˇㄖㄢˊ, ㄨㄛˇ ㄏㄞˊㄕˋ ㄅㄨˋㄋㄥˊ ㄕㄨㄛˉ。
+./custom_character_voice/linghua/processed_105.wav|ㄋㄢˊㄕㄨㄞˋ, ㄓㄣˉㄉㄜ˙ ㄕˋ ㄧˉㄐㄧㄢˋ ㄏㄣˇㄋㄢˊ ㄉㄜ˙ ㄕˋㄑㄧㄥˊ。
+./custom_character_voice/linghua/processed_106.wav|ㄨㄛˇ ㄅㄧˋㄒㄩˉ ㄅㄨˋㄉㄨㄢˋ ㄍㄠˋㄐㄧㄝˋ ㄗˋㄐㄧˇ ㄕˋ ㄕㄣˊㄌㄧˋ ㄐㄧㄚˉ ㄉㄜ˙ ㄉㄚˋ ㄒㄧㄠˇㄐㄧㄝˇ ㄗㄞˋ ㄨˊㄕㄨˋㄕㄨㄤˉ ㄧㄢˇㄐㄧㄥˉ， ㄨˊㄕㄨˋ ㄖㄣˊ ㄉㄜ˙ ㄑㄧˉㄆㄢˋ ㄓˉㄒㄧㄚˋ ㄅㄞˇ ㄔㄨˉ ㄨㄢˊㄇㄟˇㄨˊㄑㄩㄝˉ ㄉㄜ˙ ㄗˉㄊㄞˋ。
+./custom_character_voice/linghua/processed_107.wav|ㄓㄜˋㄧㄤˋ ㄉㄜ˙ ㄨㄛˇ, ㄕˋㄈㄡˇ ㄧㄝˇ ㄍㄞˉ ㄓㄨㄟˉㄑㄧㄡˊ ㄗˋㄐㄧˇ ㄉㄜ˙ ㄩㄢˋㄨㄤˋ ㄋㄜ˙?
+./custom_character_voice/linghua/processed_108.wav|ㄕˋㄈㄡˇ ㄧㄝˇ ㄍㄞˉ… ㄖㄤˋ ㄋㄧˇ ㄌㄧˇㄐㄧㄝˇ ㄨㄛˇ ㄉㄜ˙ ㄒㄧㄣˉㄧˋ ㄋㄜ˙?
+./custom_character_voice/linghua/processed_109.wav|ㄨㄛˇ ㄉㄨㄟˋ ㄧˋㄍㄨㄛˊ ㄌㄧㄠˋㄌㄧˇ ㄏㄣˇ ㄧㄡˇ ㄒㄧㄥˋㄑㄩˋ, ㄙㄨㄟˉㄖㄢˊ ㄏㄣˇㄕㄠˇ ㄧㄡˇ ㄔㄤˊㄕˋ ㄉㄜ˙ ㄐㄧˉㄏㄨㄟˋ。
+./custom_character_voice/linghua/processed_110.wav|ㄖㄨˊㄍㄨㄛˇ ㄕˋ ㄧㄠˋ ㄐㄩˊㄒㄧㄢˋ ㄗㄞˋ ㄉㄠˋㄑㄧˉ ㄌㄧㄠˋㄌㄧˇ ㄋㄟˋ ㄉㄜ˙ ㄏㄨㄚˋ, ㄧㄥˉㄍㄞˉ ㄕˋ ㄔㄚˊ ㄆㄠˋ ㄈㄢˋ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_111.wav|ㄅㄨˋㄋㄥˊ ㄍㄟˇ ㄧˋㄅㄢˉ ㄎㄜˋㄖㄣˊ ㄎㄢˋㄐㄧㄢˋ。
+./custom_character_voice/linghua/processed_112.wav|ㄨㄛˇ ㄧㄝˇ ㄕˋ ㄊㄡˉㄊㄡˉ ㄍㄠˋㄙㄨˋ ㄋㄧˇ ㄉㄜ˙。
+./custom_character_voice/linghua/processed_113.wav|ㄙㄨㄟˉㄖㄢˊ ㄅㄨˊㄕˋ ㄅㄨˋㄋㄥˊ ㄔˉ, ㄉㄢˋ ㄨㄛˇ ㄉㄨㄟˋ ㄉㄨㄥˋㄨˋ ㄓˉㄈㄤˊ ㄏㄨㄛˋ ㄋㄟˋㄗㄤˋ… ㄉㄚˋㄍㄞˋ ㄏㄨㄟˋ ㄧㄡˇㄉㄧㄢˇ…
+./custom_character_voice/linghua/processed_114.wav|ㄑㄧㄥˇ ㄙㄨㄟˊ ㄨㄛˇ ㄌㄞˊ, ㄓˇㄧㄠˋ ㄗㄡˇ ㄧˉㄒㄧㄠˇ ㄉㄨㄢˋㄌㄨˋ, ㄅㄨˊㄏㄨㄟˋ ㄉㄢˉㄍㄜˉ ㄋㄧˇ ㄊㄞˋ ㄓㄤˇㄕˊㄐㄧㄢˉ ㄉㄜ˙。
+./custom_character_voice/linghua/processed_115.wav|ㄉㄚˇㄊㄧㄥˉ ㄉㄠˋ ㄋㄧˇ ㄉㄜ˙ ㄕㄥˉㄖˋ ㄓˉㄏㄡˋ, ㄨㄛˇ ㄐㄧㄡˋ ㄊㄧˊㄑㄧㄢˊ ㄌㄜ˙ ㄧˉㄉㄨㄢˋㄕˊㄐㄧㄢˉ ㄎㄞˉㄕˇ ㄔㄡˊㄅㄟˋ。
+./custom_character_voice/linghua/processed_116.wav|ㄅㄧˇㄑㄧˇ ㄇㄧㄥˊㄍㄨㄟˋ ㄉㄜ˙ ㄌㄧˇㄨˋ ㄨㄛˇ ㄨㄤˋㄗˋ ㄘㄞˉㄘㄜˋ ㄏㄨㄛˋㄒㄩˇ ㄓㄜˋㄧㄤˋ ㄉㄜ˙ ㄌㄧˇㄨˋ ㄏㄨㄟˋ ㄍㄥˋㄏㄜˊ ㄨㄛˇㄇㄣ˙ ㄉㄜ˙ ㄑㄧㄥˊㄧˋ?
+./custom_character_voice/linghua/processed_117.wav|ㄘˇㄘˋ, ㄐㄧㄡˋ ㄑㄧㄥˇ ㄖㄤˋ ㄨㄛˇ ㄧˇㄕㄢˋ ㄨˊㄨㄟˋㄌㄧˇ ㄅㄚ˙。
+./custom_character_voice/linghua/processed_118.wav|ㄕˉㄌㄧˇ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_119.wav|ㄍㄢˇㄒㄧㄝˋ, ㄩˇ ㄋㄧˇ ㄑㄧㄝˉㄘㄨㄛˉ ㄕˇ ㄨㄛˇ ㄕㄡˉㄧˋ ㄌㄧㄤˊㄉㄨㄛˉ, ㄒㄧㄤˉㄒㄧㄣˋ ㄗㄞˋ ㄐㄧㄢˋㄕㄨˋ ㄕㄤˋ ㄧㄝˇ ㄋㄥˊ ㄍㄥˋ ㄐㄧㄣˋ ㄧˉㄅㄨˋ。
+./custom_character_voice/linghua/processed_120.wav|ㄉㄨㄛˉㄎㄨㄟˉ ㄋㄧˇ ㄉㄜ˙ ㄉㄧㄢˇㄅㄛˉ。
+./custom_character_voice/linghua/processed_121.wav|ㄨㄛˇ ㄉㄨㄟˋ ㄗˋㄐㄧˇ ㄉㄜ˙ ㄋㄥˊㄌㄧˋ ㄧㄝˇ ㄌㄧˇㄐㄧㄝˇ ㄉㄜˊ ㄍㄥˋ ㄊㄡˋㄔㄜˋ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_122.wav|ㄍㄥˋㄐㄧㄚˉ ㄧㄡˊㄖㄣˋㄧㄡˇㄩˊ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_123.wav|ㄕㄡˉ ㄈㄥˋㄒㄧㄥˊ ㄕˋㄨˋ ㄓˉㄩˊ, ㄕㄣˋㄓˋ ㄧㄡˇㄎㄨㄥˋ ㄔㄤˊㄕˋ ㄧˉㄒㄧㄝˉ ㄒㄧㄣˉ ㄉㄜ˙ ㄕˋㄑㄧㄥˊ。
+./custom_character_voice/linghua/processed_124.wav|ㄒㄧㄤˇㄧㄠˋ ㄌㄞˊ ㄕˋㄕˋ ㄗˋㄐㄧˇ ㄒㄩㄝˊㄗㄨㄛˋ ㄉㄜ˙ ㄉㄧㄢˇㄒㄧㄣˉ ㄇㄚ˙?
+./custom_character_voice/linghua/processed_125.wav|ㄒㄧㄣˉ ㄧㄡˇㄙㄨㄛˇㄙˉ, ㄙˉㄕˋ ㄖㄨˊㄆㄢˋ。
+./custom_character_voice/linghua/processed_126.wav|ㄊㄨˊㄌㄧㄠˋ ㄈㄨˊㄕˋ ㄕˋ, ㄌㄧㄡˊㄓㄨˋ ㄋㄢˊ。
+./custom_character_voice/linghua/processed_127.wav|ㄅㄠˋㄑㄧㄢˋ, ㄇㄧㄥˊㄇㄧㄥˊ ㄕˋ ㄓˊㄉㄜ˙ ㄍㄠˉㄒㄧㄥˋ ㄉㄜ˙ ㄕˊㄏㄡˋ, ㄨㄛˇ ㄑㄩㄝˋ ㄒㄧㄤˇㄑㄧˇ ㄌㄜ˙ ㄋㄚˋㄇㄜ˙ ㄅㄟˉㄕㄤˉ ㄉㄜ˙ ㄕˊㄎㄜˋ……
+./custom_character_voice/linghua/processed_128.wav|ㄒㄧㄤˉㄔㄨˋ ㄉㄜ˙ ㄕˊㄐㄧㄢˉㄍㄨㄛˋ ㄩˊ ㄔㄤˋㄏㄨㄢˇ ㄐㄧㄥˋㄖㄢˊ ㄖㄤˋ ㄨㄛˇ ㄏㄞˋㄆㄚˋ ㄗㄞˋㄘˋ ㄕˉㄑㄩˋ。
+./custom_character_voice/linghua/processed_129.wav|ㄕˉ… ㄕˉㄊㄞˋ ㄌㄜ˙ ㄧㄚ˙。
+./custom_character_voice/linghua/processed_130.wav|ㄅㄧˇㄨˇ ㄅㄚ˙ ㄕˉㄌㄧˇ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_132.wav|ㄕㄣˊㄌㄧˇㄌㄧㄡˊ  ㄕㄨㄤˉㄇㄧㄝˋ。
+./custom_character_voice/linghua/processed_133.wav|ㄓㄜˋ ㄇㄧˋ ㄑㄧˊ, ㄧㄝˇ ㄙㄨㄢˋ ㄉㄜˊ ㄧˉㄓㄨㄥˇ ㄧㄚˇㄑㄩˋ。
+./custom_character_voice/linghua/processed_134.wav|ㄐㄧㄣˉㄖˋ ㄩㄣˋㄕˋ ㄅㄨˋㄘㄨㄛˋ, ㄨㄛˇㄏㄨㄟˋ ㄓㄣˉㄒㄧˉ ㄓㄜˋ ㄧˉㄈㄣˋ ㄒㄧㄥˋㄩㄣˋ, ㄅㄨˋㄖㄨㄥˊ ㄒㄧㄠˇㄑㄩˋ ㄋㄜ˙。
+./custom_character_voice/linghua/processed_135.wav|ㄨㄛˇ ㄉㄜ˙ ㄉㄨㄟˋㄕㄡˇ, ㄍㄞˉ ㄐㄩㄝˊㄉㄨㄢˋ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_136.wav|ㄕˉㄊㄞˋ ㄌㄜ˙ ㄨㄛˇ ㄏㄞˊㄧㄡˇ ㄨㄟˋㄐㄧㄣˇ ㄓˉㄕˋ ㄖㄤˋ ㄐㄧㄚˉㄗㄨˊ ㄇㄥˊㄒㄧㄡˉ ㄌㄜ˙。
+./custom_character_voice/linghua/processed_137.wav|ㄉㄜ˙ ㄐㄧㄚˉㄏㄨㄛ˙… ㄗㄣˇㄇㄜ˙ ㄏㄨㄟˋ…
+./custom_character_voice/linghua/processed_138.wav|ㄕㄣˊㄌㄧˇㄌㄧㄥˊㄏㄨㄚˊ。
+./custom_character_voice/linghua/processed_140.wav|ㄑㄧㄥˇ ㄘˋㄐㄧㄠˋ。
+./custom_character_voice/linghua/processed_141.wav|ㄑㄧㄥˇ ㄉㄨㄛˉ ㄍㄨㄢˉㄓㄠˋ。

filelists/short_character_anno.list

@@ -0,0 +1,137 @@
+./custom_character_voice/linghua/processed_0.wav|0|稻妻神里流太刀术皆传神里绫华参上
+./custom_character_voice/linghua/processed_1.wav|0|請多指教啊
+./custom_character_voice/linghua/processed_2.wav|0|這樣悠閒安穩的時光,如果再多一點就好了。
+./custom_character_voice/linghua/processed_3.wav|0|我真貪心啊
+./custom_character_voice/linghua/processed_4.wav|0|就和茶一样,细细品味,才能理解其中风雅。
+./custom_character_voice/linghua/processed_5.wav|0|都是旅行者平日里的生活吗?
+./custom_character_voice/linghua/processed_6.wav|0|感覺又多瞭解了你一些。
+./custom_character_voice/linghua/processed_7.wav|0|剑抱业名工懷寶
+./custom_character_voice/linghua/processed_8.wav|0|随我一同避雨吧
+./custom_character_voice/linghua/processed_9.wav|0|所大人,是在訴說什麼嗎?
+./custom_character_voice/linghua/processed_10.wav|0|寂银妆素,桔高映琼枝。
+./custom_character_voice/linghua/processed_11.wav|0|嗯……美景当前,只差一壶茶与之相衬呢。
+./custom_character_voice/linghua/processed_12.wav|0|要去哪邊走走嗎?
+./custom_character_voice/linghua/processed_13.wav|0|眼睛,向著風吹來的方向
+./custom_character_voice/linghua/processed_14.wav|0|哎呀,很舒服吧
+./custom_character_voice/linghua/processed_15.wav|0|旅行者
+./custom_character_voice/linghua/processed_16.wav|0|這樣在清晨見你一面,我會忍不住覺得,接下來將是順利的一天。
+./custom_character_voice/linghua/processed_18.wav|0|茶饭之后,难免略有困倦是否有兴致下盘棋提神呢?
+./custom_character_voice/linghua/processed_20.wav|0|会是一个良宵呢
+./custom_character_voice/linghua/processed_21.wav|0|知是梦 何须醒
+./custom_character_voice/linghua/processed_22.wav|0|不比真如,一相會。
+./custom_character_voice/linghua/processed_23.wav|0|稻妻幕府社奉行神里家,位於稻妻名門中的筆頭之格位。
+./custom_character_voice/linghua/processed_24.wav|0|作为三奉行之一,掌管祭祀活动与人文艺术
+./custom_character_voice/linghua/processed_25.wav|0|双亲过世之后,族内的大小事务便由兄长和我承担了。
+./custom_character_voice/linghua/processed_26.wav|0|很多人因為我是白鷺公主,是社奉行神里家的大小姐,而敬重我。
+./custom_character_voice/linghua/processed_27.wav|0|他们所敬重的,只是我所身处的地位,与绫华我是怎样的人并无关系。
+./custom_character_voice/linghua/processed_28.wav|0|我想,能真正走进我的,或许只有,如今的我
+./custom_character_voice/linghua/processed_29.wav|0|依舊想成為值得大家信任的人
+./custom_character_voice/linghua/processed_30.wav|0|鼓舞我的原因,已不再是肩上的责任,或他人的期待。
+./custom_character_voice/linghua/processed_31.wav|0|是因為……你也是這樣的人啊
+./custom_character_voice/linghua/processed_32.wav|0|如果您有空,我們一步暮露茶室如何?
+./custom_character_voice/linghua/processed_33.wav|0|在这样恬静的日子,略加交流茶艺心得,想来是颇具雅趣的。
+./custom_character_voice/linghua/processed_34.wav|0|如果有机会的话,我想试着和你共度异国的节日。
+./custom_character_voice/linghua/processed_35.wav|0|遵循的当地风俗、礼仪规范,还有伴手礼的推荐
+./custom_character_voice/linghua/processed_36.wav|0|可以麻烦你…一一指导我吗?
+./custom_character_voice/linghua/processed_37.wav|0|神之眼,即是胸怀大志之人所获的应兽
+./custom_character_voice/linghua/processed_38.wav|0|如果问我有什么志向的话
+./custom_character_voice/linghua/processed_39.wav|0|这个还是保密吧
+./custom_character_voice/linghua/processed_40.wav|0|只是一个微不足道的梦想罢了
+./custom_character_voice/linghua/processed_41.wav|0|常说 蝉茶一味
+./custom_character_voice/linghua/processed_42.wav|0|剑禅亦如
+./custom_character_voice/linghua/processed_43.wav|0|那么剑和茶,又是什么关系呢?
+./custom_character_voice/linghua/processed_44.wav|0|你在城中,見過枯萎的櫻花樹嗎?
+./custom_character_voice/linghua/processed_45.wav|0|枯枝美 讓我想到春天盛開之景
+./custom_character_voice/linghua/processed_46.wav|0|不过,别人似乎并不这么想
+./custom_character_voice/linghua/processed_47.wav|0|開花的音數會被移走
+./custom_character_voice/linghua/processed_48.wav|0|就算一次也好,真想看到它再次開放。
+./custom_character_voice/linghua/processed_49.wav|0|是我很重要的朋友
+./custom_character_voice/linghua/processed_50.wav|0|天冷和阳光,总是感染着我
+./custom_character_voice/linghua/processed_51.wav|0|意义上来说,他就像是我的另一个兄长一样。
+./custom_character_voice/linghua/processed_52.wav|0|成為了神力家的一員
+./custom_character_voice/linghua/processed_53.wav|0|领奉行工作的九条沙罗,他总是一脸严肃
+./custom_character_voice/linghua/processed_54.wav|0|她曾经起过几次争端
+./custom_character_voice/linghua/processed_55.wav|0|���本质是忠义之士,这点无可否认
+./custom_character_voice/linghua/processed_56.wav|0|苏冶岳先生
+./custom_character_voice/linghua/processed_57.wav|0|她和九条小姐一样,是天领奉行的人。
+./custom_character_voice/linghua/processed_58.wav|0|她……也是一个很有原则的人。
+./custom_character_voice/linghua/processed_59.wav|0|这些原则的坚持,甚至比九条小姐更执着。
+./custom_character_voice/linghua/processed_60.wav|0|不过,什么事应该被算在这些原则之内……我想,或许只有路也愿先生自己知道吧。
+./custom_character_voice/linghua/processed_61.wav|0|小幼稚孩子,最近有沒有給你添什麼麻煩呢?
+./custom_character_voice/linghua/processed_62.wav|0|如果看見他偷懶,可以直接告訴我。
+./custom_character_voice/linghua/processed_63.wav|0|理所當然的看法嗎?
+./custom_character_voice/linghua/processed_64.wav|0|不該由我等下屬隨意議論
+./custom_character_voice/linghua/processed_65.wav|0|哼,將軍大人他在追求永恆之路上,可能也很孤獨吧。
+./custom_character_voice/linghua/processed_66.wav|0|一刀,并且活了下来。该说是堪称伟业的经历了吧。
+./custom_character_voice/linghua/processed_67.wav|0|虽然对我来说,她是真正的神明
+./custom_character_voice/linghua/processed_68.wav|0|可以更改到期的命运
+./custom_character_voice/linghua/processed_69.wav|0|但是如果是和你起了冲突的话,我一定会站在你这一边的。
+./custom_character_voice/linghua/processed_70.wav|0|八重公司大人的合作向来很愉快
+./custom_character_voice/linghua/processed_71.wav|0|看,操辦節慶慶典非常勞神費力,營收多半也不好看。
+./custom_character_voice/linghua/processed_72.wav|0|八重公司大人操办的产业,真的既风雅又有收成。
+./custom_character_voice/linghua/processed_73.wav|0|作為神力家的家主,兄長平日需要出面諸多場合。
+./custom_character_voice/linghua/processed_74.wav|0|我雖盡力分擔兄長肩上所負擔的責任,卻也無法緩和他常年積壓在身的疲倦。
+./custom_character_voice/linghua/processed_75.wav|0|我也是当时为了妹妹着想,希望你能劝言,让兄长多多注意身体啊。
+./custom_character_voice/linghua/processed_76.wav|0|節慶慶典是聖奉行和巫女中的責任
+./custom_character_voice/linghua/processed_77.wav|0|作為煙火專家,也的確能讓氣氛熱烈起來。
+./custom_character_voice/linghua/processed_78.wav|0|合作多了,因而漸漸熟絡
+./custom_character_voice/linghua/processed_79.wav|0|治安和消防問題嗎?我們也會一併納入預先規劃的
+./custom_character_voice/linghua/processed_80.wav|0|這些原因而看不到櫻花,才會讓人們心中留下缺憾吧。
+./custom_character_voice/linghua/processed_81.wav|0|你的请求,对我来说很特别呢既然把你当作朋友,我也应坦诚以待
+./custom_character_voice/linghua/processed_82.wav|0|不过,事关神里家的秘密,还希望你能守口如瓶。
+./custom_character_voice/linghua/processed_83.wav|0|神力嘉因為沒能保護好國寶級別的刀功,遭受了不小的衝擊。
+./custom_character_voice/linghua/processed_84.wav|0|别人中阴谋算计我们折损了诸多臣下受到许多责罚
+./custom_character_voice/linghua/processed_85.wav|0|甚至因之早衰神力家在幕府中的地位也一落千丈
+./custom_character_voice/linghua/processed_86.wav|0|好在兄長繼任後力挽狂瀾
+./custom_character_voice/linghua/processed_87.wav|0|家門雖已復興,射風醒一夕也上下其心,但大事要事方面仍靠兄長定奪。
+./custom_character_voice/linghua/processed_88.wav|0|他们所传的断刀之术,也因星象、用途、旷植、炉火环境、人之性格、元素变化的不同而有所区分
+./custom_character_voice/linghua/processed_89.wav|0|是刀弓之间所说的雷电五川
+./custom_character_voice/linghua/processed_90.wav|0|藏明刀的刀工,也被算作是本領通神的神士相關人員,歸屬統籌文化、藝術、祭祀的射鳳形意派管理。
+./custom_character_voice/linghua/processed_91.wav|0|出現了刀工背叛的事情,自然就是神里家的獨辦不力了。
+./custom_character_voice/linghua/processed_92.wav|0|對我來說,母親是意義非凡的存在。
+./custom_character_voice/linghua/processed_93.wav|0|服装,优雅,无论遇到怎样的局面,都会露出沉浸的笑容,以从容不迫的态度,操持着审理家大大小小的事物。
+./custom_character_voice/linghua/processed_94.wav|0|感情是完美的化身也不為過
+./custom_character_voice/linghua/processed_95.wav|0|但自从她离世的那一刻起,我就深切地意识到,我已经不是那个可以躲在母亲身后的小灵花了。
+./custom_character_voice/linghua/processed_96.wav|0|原来要说的话,可能不太符合盗窃幕府射风行神力家的身份
+./custom_character_voice/linghua/processed_97.wav|0|不过,就我的判断,你应该能容许我小小的任性吧
+./custom_character_voice/linghua/processed_98.wav|0|稍微有点累了可以让我靠一���你的肩膀吗?
+./custom_character_voice/linghua/processed_99.wav|0|舅舅
+./custom_character_voice/linghua/processed_100.wav|0|我都很喜欢雅乐诗词棋艺与舞蹈你也感兴趣吗?
+./custom_character_voice/linghua/processed_102.wav|0|……是啊,只要生活在這個世界上,就難免遇到種種不如意的事情。
+./custom_character_voice/linghua/processed_103.wav|0|但如果是因為自己的事情,就讓周遭的人擔心的話……
+./custom_character_voice/linghua/processed_104.wav|0|果然,我還是不能說。
+./custom_character_voice/linghua/processed_105.wav|0|男帅,真的是一件很难的事情
+./custom_character_voice/linghua/processed_106.wav|0|我必须不断告诫自己是神力家的大小姐在无数双眼睛、无数人的期盼之下摆出完美无缺的姿态
+./custom_character_voice/linghua/processed_107.wav|0|这样的我,是否也该追求自己的愿望呢?
+./custom_character_voice/linghua/processed_108.wav|0|是否也該…讓你理解我的心意呢?
+./custom_character_voice/linghua/processed_109.wav|0|我对异国料理很有兴趣,虽然很少有尝试的机会。
+./custom_character_voice/linghua/processed_110.wav|0|如果是要局限在稻妻料理內的話,應該是茶泡飯吧。
+./custom_character_voice/linghua/processed_111.wav|0|不能給一般客人看見
+./custom_character_voice/linghua/processed_112.wav|0|我也是偷偷告訴你的
+./custom_character_voice/linghua/processed_113.wav|0|虽然不是不能吃,但我对动物脂肪或内脏…大概会有点…
+./custom_character_voice/linghua/processed_114.wav|0|请随我来,只要走一小段路,不会耽搁你太长时间的。
+./custom_character_voice/linghua/processed_115.wav|0|打听到你的生日之后,我就提前了一段时间开始筹备。
+./custom_character_voice/linghua/processed_116.wav|0|比起名贵的礼物我妄自猜测或许这样的礼物会更合我们的情意?
+./custom_character_voice/linghua/processed_117.wav|0|此次,就請讓我以善無為禮吧。
+./custom_character_voice/linghua/processed_118.wav|0|失禮了
+./custom_character_voice/linghua/processed_119.wav|0|感謝,與你切磋使我收益良多,相信在劍術上也能更進一步。
+./custom_character_voice/linghua/processed_120.wav|0|多亏你的点播
+./custom_character_voice/linghua/processed_121.wav|0|我對自己的能力也理解得更透徹了。
+./custom_character_voice/linghua/processed_122.wav|0|更加游刃有余了
+./custom_character_voice/linghua/processed_123.wav|0|收奉行事物之餘,甚至有空嘗試一些新的事情。
+./custom_character_voice/linghua/processed_124.wav|0|想要來試試自己學做的點心嗎?
+./custom_character_voice/linghua/processed_125.wav|0|心有所思,思事如盼。
+./custom_character_voice/linghua/processed_126.wav|0|塗料服侍世,留住難。
+./custom_character_voice/linghua/processed_127.wav|0|抱歉,明明是值得高兴的时候,我却想起了那么悲伤的时刻……
+./custom_character_voice/linghua/processed_128.wav|0|相處的時間過於暢緩竟然讓我害怕再次失去
+./custom_character_voice/linghua/processed_129.wav|0|失…失態了呀
+./custom_character_voice/linghua/processed_130.wav|0|比武吧失禮了
+./custom_character_voice/linghua/processed_132.wav|0|神里流 霜灭
+./custom_character_voice/linghua/processed_133.wav|0|這祕奇,也算得一種雅趣
+./custom_character_voice/linghua/processed_134.wav|0|今日运势不错,我会珍惜这一份幸运,不容小觑呢。
+./custom_character_voice/linghua/processed_135.wav|0|我的对手,该决断了
+./custom_character_voice/linghua/processed_136.wav|0|失态了我还有未尽之事让家族蒙羞了
+./custom_character_voice/linghua/processed_137.wav|0|的家伙…怎么会…
+./custom_character_voice/linghua/processed_138.wav|0|神里绫华
+./custom_character_voice/linghua/processed_140.wav|0|请赐教
+./custom_character_voice/linghua/processed_141.wav|0|請多關照

infer_onnx.py

@@ -0,0 +1,58 @@
+import torch
+import onnxruntime
+import numpy as np
+import argparse
+
+import commons
+import utils
+from text import text_to_sequence
+
+from scipy.io.wavfile import write
+
+
+def get_text(text, hps):
+    text_norm = text_to_sequence(text, hps.data.text_cleaners)
+    if hps.data.add_blank:
+        text_norm = commons.intersperse(text_norm, 0)
+    text_norm = torch.LongTensor(text_norm)
+    return text_norm
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--model", required=True, help="Path to model (.onnx)")
+    parser.add_argument(
+        "--config-path", required=True, help="Path to model config (.json)"
+    )
+    parser.add_argument(
+        "--output-wav-path", required=True, help="Path to write WAV file"
+    )
+    parser.add_argument("--text", required=True, type=str, help="Text to synthesize")
+    args = parser.parse_args()
+
+    sess_options = onnxruntime.SessionOptions()
+    model = onnxruntime.InferenceSession(str(args.model), sess_options=sess_options)
+
+    hps = utils.get_hparams_from_file(args.config_path)
+
+    phoneme_ids = get_text(args.text, hps)
+    text = np.expand_dims(np.array(phoneme_ids, dtype=np.int64), 0)
+    text_lengths = np.array([text.shape[1]], dtype=np.int64)
+    scales = np.array([0.667, 1.0, 0.8], dtype=np.float32)
+    sid = None
+
+    audio = model.run(
+        None,
+        {
+            "input": text,
+            "input_lengths": text_lengths,
+            "scales": scales,
+            "sid": sid,
+        },
+    )[0].squeeze((0, 1))
+
+    write(data=audio, rate=hps.data.sampling_rate, filename=args.output_wav_path)
+
+
+if __name__ == "__main__":
+    main()

inference.py

@@ -0,0 +1,63 @@
+## LJSpeech
+import torch
+
+import commons
+import utils
+from models import SynthesizerTrn
+from text.symbols import symbols
+from text import text_to_sequence
+
+from scipy.io.wavfile import write
+
+
+def get_text(text, hps):
+    text_norm = text_to_sequence(text, hps.data.text_cleaners)
+    if hps.data.add_blank:
+        text_norm = commons.intersperse(text_norm, 0)
+    text_norm = torch.LongTensor(text_norm)
+    return text_norm
+
+
+CONFIG_PATH = "./configs/vits2_ljs_nosdp.json"
+MODEL_PATH = "./logs/G_114000.pth"
+TEXT = "VITS-2 is Awesome!"
+OUTPUT_WAV_PATH = "sample_vits2.wav"
+
+hps = utils.get_hparams_from_file(CONFIG_PATH)
+
+if (
+    "use_mel_posterior_encoder" in hps.model.keys()
+    and hps.model.use_mel_posterior_encoder == True
+):
+    print("Using mel posterior encoder for VITS2")
+    posterior_channels = 80  # vits2
+    hps.data.use_mel_posterior_encoder = True
+else:
+    print("Using lin posterior encoder for VITS1")
+    posterior_channels = hps.data.filter_length // 2 + 1
+    hps.data.use_mel_posterior_encoder = False
+
+net_g = SynthesizerTrn(
+    len(symbols),
+    posterior_channels,
+    hps.train.segment_size // hps.data.hop_length,
+    **hps.model
+).cuda()
+_ = net_g.eval()
+
+_ = utils.load_checkpoint(MODEL_PATH, net_g, None)
+
+stn_tst = get_text(TEXT, hps)
+with torch.no_grad():
+    x_tst = stn_tst.cuda().unsqueeze(0)
+    x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).cuda()
+    audio = (
+        net_g.infer(
+            x_tst, x_tst_lengths, noise_scale=0.667, noise_scale_w=0.8, length_scale=1
+        )[0][0, 0]
+        .data.cpu()
+        .float()
+        .numpy()
+    )
+
+write(data=audio, rate=hps.data.sampling_rate, filename=OUTPUT_WAV_PATH)

inference_ms.py

@@ -0,0 +1,71 @@
+## VCTK
+import torch
+
+import commons
+import utils
+from models import SynthesizerTrn
+from text.symbols import symbols
+from text import text_to_sequence
+
+from scipy.io.wavfile import write
+
+
+def get_text(text, hps):
+    text_norm = text_to_sequence(text, hps.data.text_cleaners)
+    if hps.data.add_blank:
+        text_norm = commons.intersperse(text_norm, 0)
+    text_norm = torch.LongTensor(text_norm)
+    return text_norm
+
+
+CONFIG_PATH = "./configs/vits2_vctk_base.json"
+MODEL_PATH = "/path/to/pretrained_vctk.pth"
+TEXT = "VITS-2 is Awesome!"
+SPK_ID = 4
+OUTPUT_WAV_PATH = "sample_vits2_ms.wav"
+
+hps = utils.get_hparams_from_file(CONFIG_PATH)
+
+if (
+    "use_mel_posterior_encoder" in hps.model.keys()
+    and hps.model.use_mel_posterior_encoder == True
+):
+    print("Using mel posterior encoder for VITS2")
+    posterior_channels = 80  # vits2
+    hps.data.use_mel_posterior_encoder = True
+else:
+    print("Using lin posterior encoder for VITS1")
+    posterior_channels = hps.data.filter_length // 2 + 1
+    hps.data.use_mel_posterior_encoder = False
+
+net_g = SynthesizerTrn(
+    len(symbols),
+    posterior_channels,
+    hps.train.segment_size // hps.data.hop_length,
+    n_speakers=hps.data.n_speakers,
+    **hps.model
+).cuda()
+_ = net_g.eval()
+
+_ = utils.load_checkpoint(MODEL_PATH, net_g, None)
+
+stn_tst = get_text(TEXT, hps)
+with torch.no_grad():
+    x_tst = stn_tst.cuda().unsqueeze(0)
+    x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).cuda()
+    sid = torch.LongTensor([SPK_ID]).cuda()
+    audio = (
+        net_g.infer(
+            x_tst,
+            x_tst_lengths,
+            sid=sid,
+            noise_scale=0.667,
+            noise_scale_w=0.8,
+            length_scale=1,
+        )[0][0, 0]
+        .data.cpu()
+        .float()
+        .numpy()
+    )
+
+write(data=audio, rate=hps.data.sampling_rate, filename=OUTPUT_WAV_PATH)

logs/pretrained_models/README.md

@@ -0,0 +1 @@
+Please put the pretrained models under this folder | 请将预训练模型放在此文件夹下

losses.py

@@ -0,0 +1,61 @@
+import torch
+from torch.nn import functional as F
+
+import commons
+
+
+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):
+            rl = rl.float().detach()
+            gl = gl.float()
+            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):
+        dr = dr.float()
+        dg = dg.float()
+        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:
+        dg = dg.float()
+        l = torch.mean((1 - dg) ** 2)
+        gen_losses.append(l)
+        loss += l
+
+    return loss, gen_losses
+
+
+def kl_loss(z_p, logs_q, m_p, logs_p, z_mask):
+    """
+    z_p, logs_q: [b, h, t_t]
+    m_p, logs_p: [b, h, t_t]
+    """
+    z_p = z_p.float()
+    logs_q = logs_q.float()
+    m_p = m_p.float()
+    logs_p = logs_p.float()
+    z_mask = z_mask.float()
+
+    kl = logs_p - logs_q - 0.5
+    kl += 0.5 * ((z_p - m_p) ** 2) * torch.exp(-2.0 * logs_p)
+    kl = torch.sum(kl * z_mask)
+    l = kl / torch.sum(z_mask)
+    return l

mel_processing.py

@@ -0,0 +1,181 @@
+import math
+import os
+from packaging import version
+import random
+import torch
+from torch import nn
+import torch.nn.functional as F
+import torch.utils.data
+import numpy as np
+import librosa
+import librosa.util as librosa_util
+from librosa.util import normalize, pad_center, tiny
+from scipy.signal import get_window
+from scipy.io.wavfile import read
+from librosa.filters import mel as librosa_mel_fn
+
+MAX_WAV_VALUE = 32768.0
+
+
+def dynamic_range_compression_torch(x, C=1, clip_val=1e-5):
+    """
+    PARAMS
+    ------
+    C: compression factor
+    """
+    return torch.log(torch.clamp(x, min=clip_val) * C)
+
+
+def dynamic_range_decompression_torch(x, C=1):
+    """
+    PARAMS
+    ------
+    C: compression factor used to compress
+    """
+    return torch.exp(x) / C
+
+
+def spectral_normalize_torch(magnitudes):
+    output = dynamic_range_compression_torch(magnitudes)
+    return output
+
+
+def spectral_de_normalize_torch(magnitudes):
+    output = dynamic_range_decompression_torch(magnitudes)
+    return output
+
+
+mel_basis = {}
+hann_window = {}
+
+
+def spectrogram_torch(y, n_fft, sampling_rate, hop_size, win_size, center=False):
+    if torch.min(y) < -1.0:
+        print("min value is ", torch.min(y))
+    if torch.max(y) > 1.0:
+        print("max value is ", torch.max(y))
+
+    global hann_window
+    dtype_device = str(y.dtype) + "_" + str(y.device)
+    wnsize_dtype_device = str(win_size) + "_" + dtype_device
+    if wnsize_dtype_device not in hann_window:
+        hann_window[wnsize_dtype_device] = torch.hann_window(win_size).to(
+            dtype=y.dtype, device=y.device
+        )
+
+    y = torch.nn.functional.pad(
+        y.unsqueeze(1),
+        (int((n_fft - hop_size) / 2), int((n_fft - hop_size) / 2)),
+        mode="reflect",
+    )
+    y = y.squeeze(1)
+
+    if version.parse(torch.__version__) >= version.parse("2"):
+        spec = torch.stft(
+            y,
+            n_fft,
+            hop_length=hop_size,
+            win_length=win_size,
+            window=hann_window[wnsize_dtype_device],
+            center=center,
+            pad_mode="reflect",
+            normalized=False,
+            onesided=True,
+            return_complex=False,
+        )
+    else:
+        spec = torch.stft(
+            y,
+            n_fft,
+            hop_length=hop_size,
+            win_length=win_size,
+            window=hann_window[wnsize_dtype_device],
+            center=center,
+            pad_mode="reflect",
+            normalized=False,
+            onesided=True,
+        )
+
+    spec = torch.sqrt(spec.pow(2).sum(-1) + 1e-6)
+    return spec
+
+
+def spec_to_mel_torch(spec, n_fft, num_mels, sampling_rate, fmin, fmax):
+    global mel_basis
+    dtype_device = str(spec.dtype) + "_" + str(spec.device)
+    fmax_dtype_device = str(fmax) + "_" + dtype_device
+    if fmax_dtype_device not in mel_basis:
+        mel = librosa_mel_fn(
+            sr=sampling_rate, n_fft=n_fft, n_mels=num_mels, fmin=fmin, fmax=fmax
+        )
+        mel_basis[fmax_dtype_device] = torch.from_numpy(mel).to(
+            dtype=spec.dtype, device=spec.device
+        )
+    spec = torch.matmul(mel_basis[fmax_dtype_device], spec)
+    spec = spectral_normalize_torch(spec)
+    return spec
+
+
+def mel_spectrogram_torch(
+    y, n_fft, num_mels, sampling_rate, hop_size, win_size, fmin, fmax, center=False
+):
+    if torch.min(y) < -1.0:
+        print("min value is ", torch.min(y))
+    if torch.max(y) > 1.0:
+        print("max value is ", torch.max(y))
+
+    global mel_basis, hann_window
+    dtype_device = str(y.dtype) + "_" + str(y.device)
+    fmax_dtype_device = str(fmax) + "_" + dtype_device
+    wnsize_dtype_device = str(win_size) + "_" + dtype_device
+    if fmax_dtype_device not in mel_basis:
+        mel = librosa_mel_fn(
+            sr=sampling_rate, n_fft=n_fft, n_mels=num_mels, fmin=fmin, fmax=fmax
+        )
+        mel_basis[fmax_dtype_device] = torch.from_numpy(mel).to(
+            dtype=y.dtype, device=y.device
+        )
+    if wnsize_dtype_device not in hann_window:
+        hann_window[wnsize_dtype_device] = torch.hann_window(win_size).to(
+            dtype=y.dtype, device=y.device
+        )
+
+    y = torch.nn.functional.pad(
+        y.unsqueeze(1),
+        (int((n_fft - hop_size) / 2), int((n_fft - hop_size) / 2)),
+        mode="reflect",
+    )
+    y = y.squeeze(1)
+
+    if version.parse(torch.__version__) >= version.parse("2"):
+        spec = torch.stft(
+            y,
+            n_fft,
+            hop_length=hop_size,
+            win_length=win_size,
+            window=hann_window[wnsize_dtype_device],
+            center=center,
+            pad_mode="reflect",
+            normalized=False,
+            onesided=True,
+            return_complex=False,
+        )
+    else:
+        spec = torch.stft(
+            y,
+            n_fft,
+            hop_length=hop_size,
+            win_length=win_size,
+            window=hann_window[wnsize_dtype_device],
+            center=center,
+            pad_mode="reflect",
+            normalized=False,
+            onesided=True,
+        )
+
+    spec = torch.sqrt(spec.pow(2).sum(-1) + 1e-6)
+
+    spec = torch.matmul(mel_basis[fmax_dtype_device], spec)
+    spec = spectral_normalize_torch(spec)
+
+    return spec

models.py

@@ -0,0 +1,1202 @@
+import copy
+import math
+import torch
+from torch import nn
+from torch.nn import functional as F
+
+import commons
+import modules
+import attentions
+import monotonic_align
+
+from torch.nn import Conv1d, ConvTranspose1d, AvgPool1d, Conv2d
+from torch.nn.utils import weight_norm, remove_weight_norm, spectral_norm
+from commons import init_weights, get_padding
+
+AVAILABLE_FLOW_TYPES = [
+    "pre_conv",
+    "fft",
+    "mono_layer_inter_residual",
+    "mono_layer_post_residual",
+]
+
+
+class StochasticDurationPredictor(nn.Module):
+    def __init__(
+        self,
+        in_channels,
+        filter_channels,
+        kernel_size,
+        p_dropout,
+        n_flows=4,
+        gin_channels=0,
+    ):
+        super().__init__()
+        filter_channels = in_channels  # it needs to be removed from future version.
+        self.in_channels = in_channels
+        self.filter_channels = filter_channels
+        self.kernel_size = kernel_size
+        self.p_dropout = p_dropout
+        self.n_flows = n_flows
+        self.gin_channels = gin_channels
+
+        self.log_flow = modules.Log()
+        self.flows = nn.ModuleList()
+        self.flows.append(modules.ElementwiseAffine(2))
+        for i in range(n_flows):
+            self.flows.append(
+                modules.ConvFlow(2, filter_channels, kernel_size, n_layers=3)
+            )
+            self.flows.append(modules.Flip())
+
+        self.post_pre = nn.Conv1d(1, filter_channels, 1)
+        self.post_proj = nn.Conv1d(filter_channels, filter_channels, 1)
+        self.post_convs = modules.DDSConv(
+            filter_channels, kernel_size, n_layers=3, p_dropout=p_dropout
+        )
+        self.post_flows = nn.ModuleList()
+        self.post_flows.append(modules.ElementwiseAffine(2))
+        for i in range(4):
+            self.post_flows.append(
+                modules.ConvFlow(2, filter_channels, kernel_size, n_layers=3)
+            )
+            self.post_flows.append(modules.Flip())
+
+        self.pre = nn.Conv1d(in_channels, filter_channels, 1)
+        self.proj = nn.Conv1d(filter_channels, filter_channels, 1)
+        self.convs = modules.DDSConv(
+            filter_channels, kernel_size, n_layers=3, p_dropout=p_dropout
+        )
+        if gin_channels != 0:
+            self.cond = nn.Conv1d(gin_channels, filter_channels, 1)
+
+    def forward(self, x, x_mask, w=None, g=None, reverse=False, noise_scale=1.0):
+        x = torch.detach(x)
+        x = self.pre(x)
+        if g is not None:
+            g = torch.detach(g)
+            x = x + self.cond(g)
+        x = self.convs(x, x_mask)
+        x = self.proj(x) * x_mask
+
+        if not reverse:
+            flows = self.flows
+            assert w is not None
+
+            logdet_tot_q = 0
+            h_w = self.post_pre(w)
+            h_w = self.post_convs(h_w, x_mask)
+            h_w = self.post_proj(h_w) * x_mask
+            e_q = (
+                torch.randn(w.size(0), 2, w.size(2)).to(device=x.device, dtype=x.dtype)
+                * x_mask
+            )
+            z_q = e_q
+            for flow in self.post_flows:
+                z_q, logdet_q = flow(z_q, x_mask, g=(x + h_w))
+                logdet_tot_q += logdet_q
+            z_u, z1 = torch.split(z_q, [1, 1], 1)
+            u = torch.sigmoid(z_u) * x_mask
+            z0 = (w - u) * x_mask
+            logdet_tot_q += torch.sum(
+                (F.logsigmoid(z_u) + F.logsigmoid(-z_u)) * x_mask, [1, 2]
+            )
+            logq = (
+                torch.sum(-0.5 * (math.log(2 * math.pi) + (e_q**2)) * x_mask, [1, 2])
+                - logdet_tot_q
+            )
+
+            logdet_tot = 0
+            z0, logdet = self.log_flow(z0, x_mask)
+            logdet_tot += logdet
+            z = torch.cat([z0, z1], 1)
+            for flow in flows:
+                z, logdet = flow(z, x_mask, g=x, reverse=reverse)
+                logdet_tot = logdet_tot + logdet
+            nll = (
+                torch.sum(0.5 * (math.log(2 * math.pi) + (z**2)) * x_mask, [1, 2])
+                - logdet_tot
+            )
+            return nll + logq  # [b]
+        else:
+            flows = list(reversed(self.flows))
+            flows = flows[:-2] + [flows[-1]]  # remove a useless vflow
+            z = (
+                torch.randn(x.size(0), 2, x.size(2)).to(device=x.device, dtype=x.dtype)
+                * noise_scale
+            )
+            for flow in flows:
+                z = flow(z, x_mask, g=x, reverse=reverse)
+            z0, z1 = torch.split(z, [1, 1], 1)
+            logw = z0
+            return logw
+
+
+class DurationPredictor(nn.Module):
+    def __init__(
+        self, in_channels, filter_channels, kernel_size, p_dropout, gin_channels=0
+    ):
+        super().__init__()
+
+        self.in_channels = in_channels
+        self.filter_channels = filter_channels
+        self.kernel_size = kernel_size
+        self.p_dropout = p_dropout
+        self.gin_channels = gin_channels
+
+        self.drop = nn.Dropout(p_dropout)
+        self.conv_1 = nn.Conv1d(
+            in_channels, filter_channels, kernel_size, padding=kernel_size // 2
+        )
+        self.norm_1 = modules.LayerNorm(filter_channels)
+        self.conv_2 = nn.Conv1d(
+            filter_channels, filter_channels, kernel_size, padding=kernel_size // 2
+        )
+        self.norm_2 = modules.LayerNorm(filter_channels)
+        self.proj = nn.Conv1d(filter_channels, 1, 1)
+
+        if gin_channels != 0:
+            self.cond = nn.Conv1d(gin_channels, in_channels, 1)
+
+    def forward(self, x, x_mask, g=None):
+        x = torch.detach(x)
+        if g is not None:
+            g = torch.detach(g)
+            x = x + self.cond(g)
+        x = self.conv_1(x * x_mask)
+        x = torch.relu(x)
+        x = self.norm_1(x)
+        x = self.drop(x)
+        x = self.conv_2(x * x_mask)
+        x = torch.relu(x)
+        x = self.norm_2(x)
+        x = self.drop(x)
+        x = self.proj(x * x_mask)
+        return x * x_mask
+
+
+class DurationDiscriminator(nn.Module):  # vits2
+    # TODO : not using "spk conditioning" for now according to the paper.
+    # Can be a better discriminator if we use it.
+    def __init__(
+        self, in_channels, filter_channels, kernel_size, p_dropout, gin_channels=0
+    ):
+        super().__init__()
+
+        self.in_channels = in_channels
+        self.filter_channels = filter_channels
+        self.kernel_size = kernel_size
+        self.p_dropout = p_dropout
+        self.gin_channels = gin_channels
+
+        self.drop = nn.Dropout(p_dropout)
+        self.conv_1 = nn.Conv1d(
+            in_channels, filter_channels, kernel_size, padding=kernel_size // 2
+        )
+        # self.norm_1 = modules.LayerNorm(filter_channels)
+        self.conv_2 = nn.Conv1d(
+            filter_channels, filter_channels, kernel_size, padding=kernel_size // 2
+        )
+        # self.norm_2 = modules.LayerNorm(filter_channels)
+        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_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_norm_2 = modules.LayerNorm(filter_channels)
+
+        # if gin_channels != 0:
+        #   self.cond = nn.Conv1d(gin_channels, in_channels, 1)
+
+        self.output_layer = nn.Sequential(nn.Linear(filter_channels, 1), nn.Sigmoid())
+
+    def forward_probability(self, x, x_mask, dur, g=None):
+        dur = self.dur_proj(dur)
+        x = torch.cat([x, dur], dim=1)
+        x = self.pre_out_conv_1(x * x_mask)
+        # x = torch.relu(x)
+        # x = self.pre_out_norm_1(x)
+        # x = self.drop(x)
+        x = self.pre_out_conv_2(x * x_mask)
+        # x = torch.relu(x)
+        # x = self.pre_out_norm_2(x)
+        # x = self.drop(x)
+        x = x * x_mask
+        x = x.transpose(1, 2)
+        output_prob = self.output_layer(x)
+        return output_prob
+
+    def forward(self, x, x_mask, dur_r, dur_hat, g=None):
+        x = torch.detach(x)
+        # if g is not None:
+        #   g = torch.detach(g)
+        #   x = x + self.cond(g)
+        x = self.conv_1(x * x_mask)
+        # x = torch.relu(x)
+        # x = self.norm_1(x)
+        # x = self.drop(x)
+        x = self.conv_2(x * x_mask)
+        # x = torch.relu(x)
+        # x = self.norm_2(x)
+        # x = self.drop(x)
+
+        output_probs = []
+        for dur in [dur_r, dur_hat]:
+            output_prob = self.forward_probability(x, x_mask, dur, g)
+            output_probs.append(output_prob)
+
+        return output_probs
+
+
+class TextEncoder(nn.Module):
+    def __init__(
+        self,
+        n_vocab,
+        out_channels,
+        hidden_channels,
+        filter_channels,
+        n_heads,
+        n_layers,
+        kernel_size,
+        p_dropout,
+        gin_channels=0,
+    ):
+        super().__init__()
+        self.n_vocab = n_vocab
+        self.out_channels = out_channels
+        self.hidden_channels = hidden_channels
+        self.filter_channels = filter_channels
+        self.n_heads = n_heads
+        self.n_layers = n_layers
+        self.kernel_size = kernel_size
+        self.p_dropout = p_dropout
+        self.gin_channels = gin_channels
+        self.emb = nn.Embedding(n_vocab, hidden_channels)
+        nn.init.normal_(self.emb.weight, 0.0, hidden_channels**-0.5)
+
+        self.encoder = attentions.Encoder(
+            hidden_channels,
+            filter_channels,
+            n_heads,
+            n_layers,
+            kernel_size,
+            p_dropout,
+            gin_channels=self.gin_channels,
+        )
+        self.proj = nn.Conv1d(hidden_channels, out_channels * 2, 1)
+
+    def forward(self, x, x_lengths, g=None):
+        x = self.emb(x) * math.sqrt(self.hidden_channels)  # [b, t, h]
+        x = torch.transpose(x, 1, -1)  # [b, h, t]
+        x_mask = torch.unsqueeze(commons.sequence_mask(x_lengths, x.size(2)), 1).to(
+            x.dtype
+        )
+
+        x = self.encoder(x * x_mask, x_mask, g=g)
+        stats = self.proj(x) * x_mask
+
+        m, logs = torch.split(stats, self.out_channels, dim=1)
+        return x, m, logs, x_mask
+
+
+class ResidualCouplingTransformersLayer(nn.Module):  # vits2
+    def __init__(
+        self,
+        channels,
+        hidden_channels,
+        kernel_size,
+        dilation_rate,
+        n_layers,
+        p_dropout=0,
+        gin_channels=0,
+        mean_only=False,
+    ):
+        assert channels % 2 == 0, "channels should be divisible by 2"
+        super().__init__()
+        self.channels = channels
+        self.hidden_channels = hidden_channels
+        self.kernel_size = kernel_size
+        self.dilation_rate = dilation_rate
+        self.n_layers = n_layers
+        self.half_channels = channels // 2
+        self.mean_only = mean_only
+        # vits2
+        self.pre_transformer = attentions.Encoder(
+            self.half_channels,
+            self.half_channels,
+            n_heads=2,
+            n_layers=2,
+            kernel_size=3,
+            p_dropout=0.1,
+            window_size=None,
+        )
+
+        self.pre = nn.Conv1d(self.half_channels, hidden_channels, 1)
+        self.enc = modules.WN(
+            hidden_channels,
+            kernel_size,
+            dilation_rate,
+            n_layers,
+            p_dropout=p_dropout,
+            gin_channels=gin_channels,
+        )
+        # vits2
+        self.post_transformer = attentions.Encoder(
+            self.hidden_channels,
+            self.hidden_channels,
+            n_heads=2,
+            n_layers=2,
+            kernel_size=3,
+            p_dropout=0.1,
+            window_size=None,
+        )
+
+        self.post = nn.Conv1d(hidden_channels, self.half_channels * (2 - mean_only), 1)
+        self.post.weight.data.zero_()
+        self.post.bias.data.zero_()
+
+    def forward(self, x, x_mask, g=None, reverse=False):
+        x0, x1 = torch.split(x, [self.half_channels] * 2, 1)
+        x0_ = self.pre_transformer(x0 * x_mask, x_mask)  # vits2
+        x0_ = x0_ + x0  # vits2 residual connection
+        h = self.pre(x0_) * x_mask  # changed from x0 to x0_ to retain x0 for the flow
+        h = self.enc(h, x_mask, g=g)
+
+        # vits2 - (experimental;uncomment the following 2 line to use)
+        # h_ = self.post_transformer(h, x_mask)
+        # h = h + h_ #vits2 residual connection
+
+        stats = self.post(h) * x_mask
+        if not self.mean_only:
+            m, logs = torch.split(stats, [self.half_channels] * 2, 1)
+        else:
+            m = stats
+            logs = torch.zeros_like(m)
+        if not reverse:
+            x1 = m + x1 * torch.exp(logs) * x_mask
+            x = torch.cat([x0, x1], 1)
+            logdet = torch.sum(logs, [1, 2])
+            return x, logdet
+        else:
+            x1 = (x1 - m) * torch.exp(-logs) * x_mask
+            x = torch.cat([x0, x1], 1)
+            return x
+
+
+class FFTransformerCouplingLayer(nn.Module):  # vits2
+    def __init__(
+        self,
+        channels,
+        hidden_channels,
+        kernel_size,
+        n_layers,
+        n_heads,
+        p_dropout=0,
+        filter_channels=768,
+        mean_only=False,
+        gin_channels=0,
+    ):
+        assert channels % 2 == 0, "channels should be divisible by 2"
+        super().__init__()
+        self.channels = channels
+        self.hidden_channels = hidden_channels
+        self.kernel_size = kernel_size
+        self.n_layers = n_layers
+        self.half_channels = channels // 2
+        self.mean_only = mean_only
+
+        self.pre = nn.Conv1d(self.half_channels, hidden_channels, 1)
+        self.enc = attentions.FFT(
+            hidden_channels,
+            filter_channels,
+            n_heads,
+            n_layers,
+            kernel_size,
+            p_dropout,
+            isflow=True,
+            gin_channels=gin_channels,
+        )
+        self.post = nn.Conv1d(hidden_channels, self.half_channels * (2 - mean_only), 1)
+        self.post.weight.data.zero_()
+        self.post.bias.data.zero_()
+
+    def forward(self, x, x_mask, g=None, reverse=False):
+        x0, x1 = torch.split(x, [self.half_channels] * 2, 1)
+        h = self.pre(x0) * x_mask
+        h_ = self.enc(h, x_mask, g=g)
+        h = h_ + h
+        stats = self.post(h) * x_mask
+        if not self.mean_only:
+            m, logs = torch.split(stats, [self.half_channels] * 2, 1)
+        else:
+            m = stats
+            logs = torch.zeros_like(m)
+
+        if not reverse:
+            x1 = m + x1 * torch.exp(logs) * x_mask
+            x = torch.cat([x0, x1], 1)
+            logdet = torch.sum(logs, [1, 2])
+            return x, logdet
+        else:
+            x1 = (x1 - m) * torch.exp(-logs) * x_mask
+            x = torch.cat([x0, x1], 1)
+            return x
+
+
+class MonoTransformerFlowLayer(nn.Module):  # vits2
+    def __init__(
+        self,
+        channels,
+        hidden_channels,
+        mean_only=False,
+        residual_connection=False,
+        # according to VITS-2 paper fig 1B set residual_connection=True
+    ):
+        assert channels % 2 == 0, "channels should be divisible by 2"
+        super().__init__()
+        self.channels = channels
+        self.hidden_channels = hidden_channels
+        self.half_channels = channels // 2
+        self.mean_only = mean_only
+        self.residual_connection = residual_connection
+        # vits2
+        self.pre_transformer = attentions.Encoder(
+            self.half_channels,
+            self.half_channels,
+            n_heads=2,
+            n_layers=2,
+            kernel_size=3,
+            p_dropout=0.1,
+            window_size=None,
+        )
+
+        self.post = nn.Conv1d(
+            self.half_channels, self.half_channels * (2 - mean_only), 1
+        )
+        self.post.weight.data.zero_()
+        self.post.bias.data.zero_()
+
+    def forward(self, x, x_mask, g=None, reverse=False):
+        if self.residual_connection:
+            if not reverse:
+                x0, x1 = torch.split(x, [self.half_channels] * 2, 1)
+                x0_ = self.pre_transformer(x0, x_mask)  # vits2
+                stats = self.post(x0_) * x_mask
+                if not self.mean_only:
+                    m, logs = torch.split(stats, [self.half_channels] * 2, 1)
+                else:
+                    m = stats
+                    logs = torch.zeros_like(m)
+                x1 = m + x1 * torch.exp(logs) * x_mask
+                x_ = torch.cat([x0, x1], 1)
+                x = x + x_
+                logdet = torch.sum(torch.log(torch.exp(logs) + 1), [1, 2])
+                logdet = logdet + torch.log(torch.tensor(2)) * (
+                    x0.shape[1] * x0.shape[2]
+                )
+                return x, logdet
+            else:
+                x0, x1 = torch.split(x, [self.half_channels] * 2, 1)
+                x0 = x0 / 2
+                x0_ = x0 * x_mask
+                x0_ = self.pre_transformer(x0, x_mask)  # vits2
+                stats = self.post(x0_) * x_mask
+                if not self.mean_only:
+                    m, logs = torch.split(stats, [self.half_channels] * 2, 1)
+                else:
+                    m = stats
+                    logs = torch.zeros_like(m)
+                x1_ = ((x1 - m) / (1 + torch.exp(-logs))) * x_mask
+                x = torch.cat([x0, x1_], 1)
+                return x
+        else:
+            x0, x1 = torch.split(x, [self.half_channels] * 2, 1)
+            x0_ = self.pre_transformer(x0 * x_mask, x_mask)  # vits2
+            h = x0_ + x0  # vits2
+            stats = self.post(h) * x_mask
+            if not self.mean_only:
+                m, logs = torch.split(stats, [self.half_channels] * 2, 1)
+            else:
+                m = stats
+                logs = torch.zeros_like(m)
+            if not reverse:
+                x1 = m + x1 * torch.exp(logs) * x_mask
+                x_ = torch.cat([x0, x1], 1)
+                logdet = torch.sum(logs, [1, 2])
+                return x, logdet
+            else:
+                x1 = (x1 - m) * torch.exp(-logs) * x_mask
+                x = torch.cat([x0, x1], 1)
+                return x
+
+
+class ResidualCouplingTransformersBlock(nn.Module):  # vits2
+    def __init__(
+        self,
+        channels,
+        hidden_channels,
+        kernel_size,
+        dilation_rate,
+        n_layers,
+        n_flows=4,
+        gin_channels=0,
+        use_transformer_flows=False,
+        transformer_flow_type="pre_conv",
+    ):
+        super().__init__()
+        self.channels = channels
+        self.hidden_channels = hidden_channels
+        self.kernel_size = kernel_size
+        self.dilation_rate = dilation_rate
+        self.n_layers = n_layers
+        self.n_flows = n_flows
+        self.gin_channels = gin_channels
+
+        self.flows = nn.ModuleList()
+        if use_transformer_flows:
+            if transformer_flow_type == "pre_conv":
+                for i in range(n_flows):
+                    self.flows.append(
+                        ResidualCouplingTransformersLayer(
+                            channels,
+                            hidden_channels,
+                            kernel_size,
+                            dilation_rate,
+                            n_layers,
+                            gin_channels=gin_channels,
+                            mean_only=True,
+                        )
+                    )
+                    self.flows.append(modules.Flip())
+            elif transformer_flow_type == "fft":
+                for i in range(n_flows):
+                    self.flows.append(
+                        FFTransformerCouplingLayer(
+                            channels,
+                            hidden_channels,
+                            kernel_size,
+                            dilation_rate,
+                            n_layers,
+                            gin_channels=gin_channels,
+                            mean_only=True,
+                        )
+                    )
+                    self.flows.append(modules.Flip())
+            elif transformer_flow_type == "mono_layer_inter_residual":
+                for i in range(n_flows):
+                    self.flows.append(
+                        modules.ResidualCouplingLayer(
+                            channels,
+                            hidden_channels,
+                            kernel_size,
+                            dilation_rate,
+                            n_layers,
+                            gin_channels=gin_channels,
+                            mean_only=True,
+                        )
+                    )
+                    self.flows.append(modules.Flip())
+                    self.flows.append(
+                        MonoTransformerFlowLayer(
+                            channels, hidden_channels, mean_only=True
+                        )
+                    )
+            elif transformer_flow_type == "mono_layer_post_residual":
+                for i in range(n_flows):
+                    self.flows.append(
+                        modules.ResidualCouplingLayer(
+                            channels,
+                            hidden_channels,
+                            kernel_size,
+                            dilation_rate,
+                            n_layers,
+                            gin_channels=gin_channels,
+                            mean_only=True,
+                        )
+                    )
+                    self.flows.append(modules.Flip())
+                    self.flows.append(
+                        MonoTransformerFlowLayer(
+                            channels,
+                            hidden_channels,
+                            mean_only=True,
+                            residual_connection=True,
+                        )
+                    )
+        else:
+            for i in range(n_flows):
+                self.flows.append(
+                    modules.ResidualCouplingLayer(
+                        channels,
+                        hidden_channels,
+                        kernel_size,
+                        dilation_rate,
+                        n_layers,
+                        gin_channels=gin_channels,
+                        mean_only=True,
+                    )
+                )
+                self.flows.append(modules.Flip())
+
+    def forward(self, x, x_mask, g=None, reverse=False):
+        if not reverse:
+            for flow in self.flows:
+                x, _ = flow(x, x_mask, g=g, reverse=reverse)
+        else:
+            for flow in reversed(self.flows):
+                x = flow(x, x_mask, g=g, reverse=reverse)
+        return x
+
+
+class ResidualCouplingBlock(nn.Module):
+    def __init__(
+        self,
+        channels,
+        hidden_channels,
+        kernel_size,
+        dilation_rate,
+        n_layers,
+        n_flows=4,
+        gin_channels=0,
+    ):
+        super().__init__()
+        self.channels = channels
+        self.hidden_channels = hidden_channels
+        self.kernel_size = kernel_size
+        self.dilation_rate = dilation_rate
+        self.n_layers = n_layers
+        self.n_flows = n_flows
+        self.gin_channels = gin_channels
+
+        self.flows = nn.ModuleList()
+        for i in range(n_flows):
+            self.flows.append(
+                modules.ResidualCouplingLayer(
+                    channels,
+                    hidden_channels,
+                    kernel_size,
+                    dilation_rate,
+                    n_layers,
+                    gin_channels=gin_channels,
+                    mean_only=True,
+                )
+            )
+            self.flows.append(modules.Flip())
+
+    def forward(self, x, x_mask, g=None, reverse=False):
+        if not reverse:
+            for flow in self.flows:
+                x, _ = flow(x, x_mask, g=g, reverse=reverse)
+        else:
+            for flow in reversed(self.flows):
+                x = flow(x, x_mask, g=g, reverse=reverse)
+        return x
+
+
+class PosteriorEncoder(nn.Module):
+    def __init__(
+        self,
+        in_channels,
+        out_channels,
+        hidden_channels,
+        kernel_size,
+        dilation_rate,
+        n_layers,
+        gin_channels=0,
+    ):
+        super().__init__()
+        self.in_channels = in_channels
+        self.out_channels = out_channels
+        self.hidden_channels = hidden_channels
+        self.kernel_size = kernel_size
+        self.dilation_rate = dilation_rate
+        self.n_layers = n_layers
+        self.gin_channels = gin_channels
+
+        self.pre = nn.Conv1d(in_channels, hidden_channels, 1)
+        self.enc = modules.WN(
+            hidden_channels,
+            kernel_size,
+            dilation_rate,
+            n_layers,
+            gin_channels=gin_channels,
+        )
+        self.proj = nn.Conv1d(hidden_channels, out_channels * 2, 1)
+
+    def forward(self, x, x_lengths, g=None):
+        x_mask = torch.unsqueeze(commons.sequence_mask(x_lengths, x.size(2)), 1).to(
+            x.dtype
+        )
+        x = self.pre(x) * x_mask
+        x = self.enc(x, x_mask, g=g)
+        stats = self.proj(x) * x_mask
+        m, logs = torch.split(stats, self.out_channels, dim=1)
+        z = (m + torch.randn_like(m) * torch.exp(logs)) * x_mask
+        return z, m, logs, x_mask
+
+
+class Generator(torch.nn.Module):
+    def __init__(
+        self,
+        initial_channel,
+        resblock,
+        resblock_kernel_sizes,
+        resblock_dilation_sizes,
+        upsample_rates,
+        upsample_initial_channel,
+        upsample_kernel_sizes,
+        gin_channels=0,
+    ):
+        super(Generator, self).__init__()
+        self.num_kernels = len(resblock_kernel_sizes)
+        self.num_upsamples = len(upsample_rates)
+        self.conv_pre = Conv1d(
+            initial_channel, upsample_initial_channel, 7, 1, padding=3
+        )
+        resblock = modules.ResBlock1 if resblock == "1" else modules.ResBlock2
+
+        self.ups = nn.ModuleList()
+        for i, (u, k) in enumerate(zip(upsample_rates, 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,
+                    )
+                )
+            )
+
+        self.resblocks = nn.ModuleList()
+        for i in range(len(self.ups)):
+            ch = upsample_initial_channel // (2 ** (i + 1))
+            for j, (k, d) in enumerate(
+                zip(resblock_kernel_sizes, resblock_dilation_sizes)
+            ):
+                self.resblocks.append(resblock(ch, k, d))
+
+        self.conv_post = Conv1d(ch, 1, 7, 1, padding=3, bias=False)
+        self.ups.apply(init_weights)
+
+        if gin_channels != 0:
+            self.cond = nn.Conv1d(gin_channels, upsample_initial_channel, 1)
+
+    def forward(self, x, g=None):
+        x = self.conv_pre(x)
+        if g is not None:
+            x = x + self.cond(g)
+
+        for i in range(self.num_upsamples):
+            x = F.leaky_relu(x, modules.LRELU_SLOPE)
+            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.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()
+
+
+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
+        self.use_spectral_norm = use_spectral_norm
+        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(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        32,
+                        128,
+                        (kernel_size, 1),
+                        (stride, 1),
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        128,
+                        512,
+                        (kernel_size, 1),
+                        (stride, 1),
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        512,
+                        1024,
+                        (kernel_size, 1),
+                        (stride, 1),
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        1024,
+                        1024,
+                        (kernel_size, 1),
+                        1,
+                        padding=(get_padding(kernel_size, 1), 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, modules.LRELU_SLOPE)
+            fmap.append(x)
+        x = self.conv_post(x)
+        fmap.append(x)
+        x = torch.flatten(x, 1, -1)
+
+        return x, fmap
+
+
+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, 16, 15, 1, padding=7)),
+                norm_f(Conv1d(16, 64, 41, 4, groups=4, padding=20)),
+                norm_f(Conv1d(64, 256, 41, 4, groups=16, padding=20)),
+                norm_f(Conv1d(256, 1024, 41, 4, groups=64, padding=20)),
+                norm_f(Conv1d(1024, 1024, 41, 4, groups=256, 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, modules.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, use_spectral_norm=False):
+        super(MultiPeriodDiscriminator, self).__init__()
+        periods = [2, 3, 5, 7, 11]
+
+        discs = [DiscriminatorS(use_spectral_norm=use_spectral_norm)]
+        discs = discs + [
+            DiscriminatorP(i, use_spectral_norm=use_spectral_norm) for i in periods
+        ]
+        self.discriminators = nn.ModuleList(discs)
+
+    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)
+            y_d_gs.append(y_d_g)
+            fmap_rs.append(fmap_r)
+            fmap_gs.append(fmap_g)
+
+        return y_d_rs, y_d_gs, fmap_rs, fmap_gs
+
+
+class SynthesizerTrn(nn.Module):
+    """
+    Synthesizer for Training
+    """
+
+    def __init__(
+        self,
+        n_vocab,
+        spec_channels,
+        segment_size,
+        inter_channels,
+        hidden_channels,
+        filter_channels,
+        n_heads,
+        n_layers,
+        kernel_size,
+        p_dropout,
+        resblock,
+        resblock_kernel_sizes,
+        resblock_dilation_sizes,
+        upsample_rates,
+        upsample_initial_channel,
+        upsample_kernel_sizes,
+        n_speakers=0,
+        gin_channels=0,
+        use_sdp=True,
+        **kwargs,
+    ):
+        super().__init__()
+        self.n_vocab = n_vocab
+        self.spec_channels = spec_channels
+        self.inter_channels = inter_channels
+        self.hidden_channels = hidden_channels
+        self.filter_channels = filter_channels
+        self.n_heads = n_heads
+        self.n_layers = n_layers
+        self.kernel_size = kernel_size
+        self.p_dropout = p_dropout
+        self.resblock = resblock
+        self.resblock_kernel_sizes = resblock_kernel_sizes
+        self.resblock_dilation_sizes = resblock_dilation_sizes
+        self.upsample_rates = upsample_rates
+        self.upsample_initial_channel = upsample_initial_channel
+        self.upsample_kernel_sizes = upsample_kernel_sizes
+        self.segment_size = segment_size
+        self.n_speakers = n_speakers
+        self.gin_channels = gin_channels
+        self.use_spk_conditioned_encoder = kwargs.get(
+            "use_spk_conditioned_encoder", False
+        )
+        self.use_transformer_flows = kwargs.get("use_transformer_flows", False)
+        self.transformer_flow_type = kwargs.get(
+            "transformer_flow_type", "mono_layer_post_residual"
+        )
+        if self.use_transformer_flows:
+            assert (
+                self.transformer_flow_type in AVAILABLE_FLOW_TYPES
+            ), f"transformer_flow_type must be one of {AVAILABLE_FLOW_TYPES}"
+        self.use_sdp = use_sdp
+        # self.use_duration_discriminator = kwargs.get("use_duration_discriminator", False)
+        self.use_noise_scaled_mas = kwargs.get("use_noise_scaled_mas", False)
+        self.mas_noise_scale_initial = kwargs.get("mas_noise_scale_initial", 0.01)
+        self.noise_scale_delta = kwargs.get("noise_scale_delta", 2e-6)
+
+        self.current_mas_noise_scale = self.mas_noise_scale_initial
+        if self.use_spk_conditioned_encoder and gin_channels > 0:
+            self.enc_gin_channels = gin_channels
+        else:
+            self.enc_gin_channels = 0
+        self.enc_p = TextEncoder(
+            n_vocab,
+            inter_channels,
+            hidden_channels,
+            filter_channels,
+            n_heads,
+            n_layers,
+            kernel_size,
+            p_dropout,
+            gin_channels=self.enc_gin_channels,
+        )
+
+        self.dec = Generator(
+            inter_channels,
+            resblock,
+            resblock_kernel_sizes,
+            resblock_dilation_sizes,
+            upsample_rates,
+            upsample_initial_channel,
+            upsample_kernel_sizes,
+            gin_channels=gin_channels,
+        )
+        self.enc_q = PosteriorEncoder(
+            spec_channels,
+            inter_channels,
+            hidden_channels,
+            5,
+            1,
+            16,
+            gin_channels=gin_channels,
+        )
+        # self.flow = ResidualCouplingBlock(inter_channels, hidden_channels, 5, 1, 4, gin_channels=gin_channels)
+        self.flow = ResidualCouplingTransformersBlock(
+            inter_channels,
+            hidden_channels,
+            5,
+            1,
+            4,
+            gin_channels=gin_channels,
+            use_transformer_flows=self.use_transformer_flows,
+            transformer_flow_type=self.transformer_flow_type,
+        )
+
+        if use_sdp:
+            self.dp = StochasticDurationPredictor(
+                hidden_channels, 192, 3, 0.5, 4, gin_channels=gin_channels
+            )
+        else:
+            self.dp = DurationPredictor(
+                hidden_channels, 256, 3, 0.5, gin_channels=gin_channels
+            )
+
+        if n_speakers > 1:
+            self.emb_g = nn.Embedding(n_speakers, gin_channels)
+
+    def forward(self, x, x_lengths, y, y_lengths, sid=None):
+        if self.n_speakers > 0:
+            g = self.emb_g(sid).unsqueeze(-1)  # [b, h, 1]
+        else:
+            g = None
+
+        x, m_p, logs_p, x_mask = self.enc_p(x, x_lengths, g=g)
+        z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=g)
+        z_p = self.flow(z, y_mask, g=g)
+
+        with torch.no_grad():
+            # negative cross-entropy
+            s_p_sq_r = torch.exp(-2 * logs_p)  # [b, d, t]
+            neg_cent1 = torch.sum(
+                -0.5 * math.log(2 * math.pi) - logs_p, [1], keepdim=True
+            )  # [b, 1, t_s]
+            neg_cent2 = torch.matmul(
+                -0.5 * (z_p**2).transpose(1, 2), s_p_sq_r
+            )  # [b, t_t, d] x [b, d, t_s] = [b, t_t, t_s]
+            neg_cent3 = torch.matmul(
+                z_p.transpose(1, 2), (m_p * s_p_sq_r)
+            )  # [b, t_t, d] x [b, d, t_s] = [b, t_t, t_s]
+            neg_cent4 = torch.sum(
+                -0.5 * (m_p**2) * s_p_sq_r, [1], keepdim=True
+            )  # [b, 1, t_s]
+            neg_cent = neg_cent1 + neg_cent2 + neg_cent3 + neg_cent4
+
+            if self.use_noise_scaled_mas:
+                epsilon = (
+                    torch.std(neg_cent)
+                    * torch.randn_like(neg_cent)
+                    * self.current_mas_noise_scale
+                )
+                neg_cent = neg_cent + epsilon
+
+            attn_mask = torch.unsqueeze(x_mask, 2) * torch.unsqueeze(y_mask, -1)
+            attn = (
+                monotonic_align.maximum_path(neg_cent, attn_mask.squeeze(1))
+                .unsqueeze(1)
+                .detach()
+            )
+
+        w = attn.sum(2)
+        if self.use_sdp:
+            l_length = self.dp(x, x_mask, w, g=g)
+            l_length = l_length / torch.sum(x_mask)
+            logw = self.dp(x, x_mask, g=g, reverse=True, noise_scale=1.0)
+            logw_ = torch.log(w + 1e-6) * x_mask
+        else:
+            logw_ = torch.log(w + 1e-6) * x_mask
+            logw = self.dp(x, x_mask, g=g)
+            l_length = torch.sum((logw - logw_) ** 2, [1, 2]) / torch.sum(
+                x_mask
+            )  # for averaging
+
+        # expand prior
+        m_p = torch.matmul(attn.squeeze(1), m_p.transpose(1, 2)).transpose(1, 2)
+        logs_p = torch.matmul(attn.squeeze(1), logs_p.transpose(1, 2)).transpose(1, 2)
+
+        z_slice, ids_slice = commons.rand_slice_segments(
+            z, y_lengths, self.segment_size
+        )
+        o = self.dec(z_slice, g=g)
+        return (
+            o,
+            l_length,
+            attn,
+            ids_slice,
+            x_mask,
+            y_mask,
+            (z, z_p, m_p, logs_p, m_q, logs_q),
+            (x, logw, logw_),
+        )
+
+    def infer(
+        self,
+        x,
+        x_lengths,
+        sid=None,
+        noise_scale=1,
+        length_scale=1,
+        noise_scale_w=1.0,
+        max_len=None,
+    ):
+        if self.n_speakers > 0:
+            g = self.emb_g(sid).unsqueeze(-1)  # [b, h, 1]
+        else:
+            g = None
+        x, m_p, logs_p, x_mask = self.enc_p(x, x_lengths, g=g)
+        if self.use_sdp:
+            logw = self.dp(x, x_mask, g=g, reverse=True, noise_scale=noise_scale_w)
+        else:
+            logw = self.dp(x, x_mask, g=g)
+        w = torch.exp(logw) * x_mask * length_scale
+        w_ceil = torch.ceil(w)
+        y_lengths = torch.clamp_min(torch.sum(w_ceil, [1, 2]), 1).long()
+        y_mask = torch.unsqueeze(commons.sequence_mask(y_lengths, None), 1).to(
+            x_mask.dtype
+        )
+        attn_mask = torch.unsqueeze(x_mask, 2) * torch.unsqueeze(y_mask, -1)
+        attn = commons.generate_path(w_ceil, attn_mask)
+
+        m_p = torch.matmul(attn.squeeze(1), m_p.transpose(1, 2)).transpose(
+            1, 2
+        )  # [b, t', t], [b, t, d] -> [b, d, t']
+        logs_p = torch.matmul(attn.squeeze(1), logs_p.transpose(1, 2)).transpose(
+            1, 2
+        )  # [b, t', t], [b, t, d] -> [b, d, t']
+
+        z_p = m_p + torch.randn_like(m_p) * torch.exp(logs_p) * noise_scale
+        z = self.flow(z_p, y_mask, g=g, reverse=True)
+        o = self.dec((z * y_mask)[:, :, :max_len], g=g)
+        return o, attn, y_mask, (z, z_p, m_p, logs_p)
+
+    ## currently vits-2 is not capable of voice conversion
+    # def voice_conversion(self, y, y_lengths, sid_src, sid_tgt):
+    #   assert self.n_speakers > 0, "n_speakers have to be larger than 0."
+    #   g_src = self.emb_g(sid_src).unsqueeze(-1)
+    #   g_tgt = self.emb_g(sid_tgt).unsqueeze(-1)
+    #   z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=g_src)
+    #   z_p = self.flow(z, y_mask, g=g_src)
+    #   z_hat = self.flow(z_p, y_mask, g=g_tgt, reverse=True)
+    #   o_hat = self.dec(z_hat * y_mask, g=g_tgt)
+    #   return o_hat, y_mask, (z, z_p, z_hat)

modules.py

@@ -0,0 +1,519 @@
+import copy
+import math
+import numpy as np
+import scipy
+import torch
+from torch import nn
+from torch.nn import functional as F
+
+from torch.nn import Conv1d, ConvTranspose1d, AvgPool1d, Conv2d
+from torch.nn.utils import weight_norm, remove_weight_norm
+
+import commons
+from commons import init_weights, get_padding
+from transforms import piecewise_rational_quadratic_transform
+
+
+LRELU_SLOPE = 0.1
+
+
+class LayerNorm(nn.Module):
+    def __init__(self, channels, eps=1e-5):
+        super().__init__()
+        self.channels = channels
+        self.eps = eps
+
+        self.gamma = nn.Parameter(torch.ones(channels))
+        self.beta = nn.Parameter(torch.zeros(channels))
+
+    def forward(self, x):
+        x = x.transpose(1, -1)
+        x = F.layer_norm(x, (self.channels,), self.gamma, self.beta, self.eps)
+        return x.transpose(1, -1)
+
+
+class ConvReluNorm(nn.Module):
+    def __init__(
+        self,
+        in_channels,
+        hidden_channels,
+        out_channels,
+        kernel_size,
+        n_layers,
+        p_dropout,
+    ):
+        super().__init__()
+        self.in_channels = in_channels
+        self.hidden_channels = hidden_channels
+        self.out_channels = out_channels
+        self.kernel_size = kernel_size
+        self.n_layers = n_layers
+        self.p_dropout = p_dropout
+        assert n_layers > 1, "Number of layers should be larger than 0."
+
+        self.conv_layers = nn.ModuleList()
+        self.norm_layers = nn.ModuleList()
+        self.conv_layers.append(
+            nn.Conv1d(
+                in_channels, hidden_channels, kernel_size, padding=kernel_size // 2
+            )
+        )
+        self.norm_layers.append(LayerNorm(hidden_channels))
+        self.relu_drop = nn.Sequential(nn.ReLU(), nn.Dropout(p_dropout))
+        for _ in range(n_layers - 1):
+            self.conv_layers.append(
+                nn.Conv1d(
+                    hidden_channels,
+                    hidden_channels,
+                    kernel_size,
+                    padding=kernel_size // 2,
+                )
+            )
+            self.norm_layers.append(LayerNorm(hidden_channels))
+        self.proj = nn.Conv1d(hidden_channels, out_channels, 1)
+        self.proj.weight.data.zero_()
+        self.proj.bias.data.zero_()
+
+    def forward(self, x, x_mask):
+        x_org = x
+        for i in range(self.n_layers):
+            x = self.conv_layers[i](x * x_mask)
+            x = self.norm_layers[i](x)
+            x = self.relu_drop(x)
+        x = x_org + self.proj(x)
+        return x * x_mask
+
+
+class DDSConv(nn.Module):
+    """
+    Dialted and Depth-Separable Convolution
+    """
+
+    def __init__(self, channels, kernel_size, n_layers, p_dropout=0.0):
+        super().__init__()
+        self.channels = channels
+        self.kernel_size = kernel_size
+        self.n_layers = n_layers
+        self.p_dropout = p_dropout
+
+        self.drop = nn.Dropout(p_dropout)
+        self.convs_sep = nn.ModuleList()
+        self.convs_1x1 = nn.ModuleList()
+        self.norms_1 = nn.ModuleList()
+        self.norms_2 = nn.ModuleList()
+        for i in range(n_layers):
+            dilation = kernel_size**i
+            padding = (kernel_size * dilation - dilation) // 2
+            self.convs_sep.append(
+                nn.Conv1d(
+                    channels,
+                    channels,
+                    kernel_size,
+                    groups=channels,
+                    dilation=dilation,
+                    padding=padding,
+                )
+            )
+            self.convs_1x1.append(nn.Conv1d(channels, channels, 1))
+            self.norms_1.append(LayerNorm(channels))
+            self.norms_2.append(LayerNorm(channels))
+
+    def forward(self, x, x_mask, g=None):
+        if g is not None:
+            x = x + g
+        for i in range(self.n_layers):
+            y = self.convs_sep[i](x * x_mask)
+            y = self.norms_1[i](y)
+            y = F.gelu(y)
+            y = self.convs_1x1[i](y)
+            y = self.norms_2[i](y)
+            y = F.gelu(y)
+            y = self.drop(y)
+            x = x + y
+        return x * x_mask
+
+
+class WN(torch.nn.Module):
+    def __init__(
+        self,
+        hidden_channels,
+        kernel_size,
+        dilation_rate,
+        n_layers,
+        gin_channels=0,
+        p_dropout=0,
+    ):
+        super(WN, self).__init__()
+        assert kernel_size % 2 == 1
+        self.hidden_channels = hidden_channels
+        self.kernel_size = (kernel_size,)
+        self.dilation_rate = dilation_rate
+        self.n_layers = n_layers
+        self.gin_channels = gin_channels
+        self.p_dropout = p_dropout
+
+        self.in_layers = torch.nn.ModuleList()
+        self.res_skip_layers = torch.nn.ModuleList()
+        self.drop = nn.Dropout(p_dropout)
+
+        if gin_channels != 0:
+            cond_layer = torch.nn.Conv1d(
+                gin_channels, 2 * hidden_channels * n_layers, 1
+            )
+            self.cond_layer = torch.nn.utils.weight_norm(cond_layer, name="weight")
+
+        for i in range(n_layers):
+            dilation = dilation_rate**i
+            padding = int((kernel_size * dilation - dilation) / 2)
+            in_layer = torch.nn.Conv1d(
+                hidden_channels,
+                2 * hidden_channels,
+                kernel_size,
+                dilation=dilation,
+                padding=padding,
+            )
+            in_layer = torch.nn.utils.weight_norm(in_layer, name="weight")
+            self.in_layers.append(in_layer)
+
+            # last one is not necessary
+            if i < n_layers - 1:
+                res_skip_channels = 2 * hidden_channels
+            else:
+                res_skip_channels = hidden_channels
+
+            res_skip_layer = torch.nn.Conv1d(hidden_channels, res_skip_channels, 1)
+            res_skip_layer = torch.nn.utils.weight_norm(res_skip_layer, name="weight")
+            self.res_skip_layers.append(res_skip_layer)
+
+    def forward(self, x, x_mask, g=None, **kwargs):
+        output = torch.zeros_like(x)
+        n_channels_tensor = torch.IntTensor([self.hidden_channels])
+
+        if g is not None:
+            g = self.cond_layer(g)
+
+        for i in range(self.n_layers):
+            x_in = self.in_layers[i](x)
+            if g is not None:
+                cond_offset = i * 2 * self.hidden_channels
+                g_l = g[:, cond_offset : cond_offset + 2 * self.hidden_channels, :]
+            else:
+                g_l = torch.zeros_like(x_in)
+
+            acts = commons.fused_add_tanh_sigmoid_multiply(x_in, g_l, n_channels_tensor)
+            acts = self.drop(acts)
+
+            res_skip_acts = self.res_skip_layers[i](acts)
+            if i < self.n_layers - 1:
+                res_acts = res_skip_acts[:, : self.hidden_channels, :]
+                x = (x + res_acts) * x_mask
+                output = output + res_skip_acts[:, self.hidden_channels :, :]
+            else:
+                output = output + res_skip_acts
+        return output * x_mask
+
+    def remove_weight_norm(self):
+        if self.gin_channels != 0:
+            torch.nn.utils.remove_weight_norm(self.cond_layer)
+        for l in self.in_layers:
+            torch.nn.utils.remove_weight_norm(l)
+        for l in self.res_skip_layers:
+            torch.nn.utils.remove_weight_norm(l)
+
+
+class ResBlock1(torch.nn.Module):
+    def __init__(self, channels, kernel_size=3, dilation=(1, 3, 5)):
+        super(ResBlock1, self).__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.convs1.apply(init_weights)
+
+        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),
+                    )
+                ),
+            ]
+        )
+        self.convs2.apply(init_weights)
+
+    def forward(self, x, x_mask=None):
+        for c1, c2 in zip(self.convs1, self.convs2):
+            xt = F.leaky_relu(x, LRELU_SLOPE)
+            if x_mask is not None:
+                xt = xt * x_mask
+            xt = c1(xt)
+            xt = F.leaky_relu(xt, LRELU_SLOPE)
+            if x_mask is not None:
+                xt = xt * x_mask
+            xt = c2(xt)
+            x = xt + x
+        if x_mask is not None:
+            x = x * x_mask
+        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, channels, kernel_size=3, dilation=(1, 3)):
+        super(ResBlock2, self).__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]),
+                    )
+                ),
+            ]
+        )
+        self.convs.apply(init_weights)
+
+    def forward(self, x, x_mask=None):
+        for c in self.convs:
+            xt = F.leaky_relu(x, LRELU_SLOPE)
+            if x_mask is not None:
+                xt = xt * x_mask
+            xt = c(xt)
+            x = xt + x
+        if x_mask is not None:
+            x = x * x_mask
+        return x
+
+    def remove_weight_norm(self):
+        for l in self.convs:
+            remove_weight_norm(l)
+
+
+class Log(nn.Module):
+    def forward(self, x, x_mask, reverse=False, **kwargs):
+        if not reverse:
+            y = torch.log(torch.clamp_min(x, 1e-5)) * x_mask
+            logdet = torch.sum(-y, [1, 2])
+            return y, logdet
+        else:
+            x = torch.exp(x) * x_mask
+            return x
+
+
+class Flip(nn.Module):
+    def forward(self, x, *args, reverse=False, **kwargs):
+        x = torch.flip(x, [1])
+        if not reverse:
+            logdet = torch.zeros(x.size(0)).to(dtype=x.dtype, device=x.device)
+            return x, logdet
+        else:
+            return x
+
+
+class ElementwiseAffine(nn.Module):
+    def __init__(self, channels):
+        super().__init__()
+        self.channels = channels
+        self.m = nn.Parameter(torch.zeros(channels, 1))
+        self.logs = nn.Parameter(torch.zeros(channels, 1))
+
+    def forward(self, x, x_mask, reverse=False, **kwargs):
+        if not reverse:
+            y = self.m + torch.exp(self.logs) * x
+            y = y * x_mask
+            logdet = torch.sum(self.logs * x_mask, [1, 2])
+            return y, logdet
+        else:
+            x = (x - self.m) * torch.exp(-self.logs) * x_mask
+            return x
+
+
+class ResidualCouplingLayer(nn.Module):
+    def __init__(
+        self,
+        channels,
+        hidden_channels,
+        kernel_size,
+        dilation_rate,
+        n_layers,
+        p_dropout=0,
+        gin_channels=0,
+        mean_only=False,
+    ):
+        assert channels % 2 == 0, "channels should be divisible by 2"
+        super().__init__()
+        self.channels = channels
+        self.hidden_channels = hidden_channels
+        self.kernel_size = kernel_size
+        self.dilation_rate = dilation_rate
+        self.n_layers = n_layers
+        self.half_channels = channels // 2
+        self.mean_only = mean_only
+
+        self.pre = nn.Conv1d(self.half_channels, hidden_channels, 1)
+        self.enc = WN(
+            hidden_channels,
+            kernel_size,
+            dilation_rate,
+            n_layers,
+            p_dropout=p_dropout,
+            gin_channels=gin_channels,
+        )
+        self.post = nn.Conv1d(hidden_channels, self.half_channels * (2 - mean_only), 1)
+        self.post.weight.data.zero_()
+        self.post.bias.data.zero_()
+
+    def forward(self, x, x_mask, g=None, reverse=False):
+        x0, x1 = torch.split(x, [self.half_channels] * 2, 1)
+        h = self.pre(x0) * x_mask
+        h = self.enc(h, x_mask, g=g)
+        stats = self.post(h) * x_mask
+        if not self.mean_only:
+            m, logs = torch.split(stats, [self.half_channels] * 2, 1)
+        else:
+            m = stats
+            logs = torch.zeros_like(m)
+
+        if not reverse:
+            x1 = m + x1 * torch.exp(logs) * x_mask
+            x = torch.cat([x0, x1], 1)
+            logdet = torch.sum(logs, [1, 2])
+            return x, logdet
+        else:
+            x1 = (x1 - m) * torch.exp(-logs) * x_mask
+            x = torch.cat([x0, x1], 1)
+            return x
+
+
+class ConvFlow(nn.Module):
+    def __init__(
+        self,
+        in_channels,
+        filter_channels,
+        kernel_size,
+        n_layers,
+        num_bins=10,
+        tail_bound=5.0,
+    ):
+        super().__init__()
+        self.in_channels = in_channels
+        self.filter_channels = filter_channels
+        self.kernel_size = kernel_size
+        self.n_layers = n_layers
+        self.num_bins = num_bins
+        self.tail_bound = tail_bound
+        self.half_channels = in_channels // 2
+
+        self.pre = nn.Conv1d(self.half_channels, filter_channels, 1)
+        self.convs = DDSConv(filter_channels, kernel_size, n_layers, p_dropout=0.0)
+        self.proj = nn.Conv1d(
+            filter_channels, self.half_channels * (num_bins * 3 - 1), 1
+        )
+        self.proj.weight.data.zero_()
+        self.proj.bias.data.zero_()
+
+    def forward(self, x, x_mask, g=None, reverse=False):
+        x0, x1 = torch.split(x, [self.half_channels] * 2, 1)
+        h = self.pre(x0)
+        h = self.convs(h, x_mask, g=g)
+        h = self.proj(h) * x_mask
+
+        b, c, t = x0.shape
+        h = h.reshape(b, c, -1, t).permute(0, 1, 3, 2)  # [b, cx?, t] -> [b, c, t, ?]
+
+        unnormalized_widths = h[..., : self.num_bins] / math.sqrt(self.filter_channels)
+        unnormalized_heights = h[..., self.num_bins : 2 * self.num_bins] / math.sqrt(
+            self.filter_channels
+        )
+        unnormalized_derivatives = h[..., 2 * self.num_bins :]
+
+        x1, logabsdet = piecewise_rational_quadratic_transform(
+            x1,
+            unnormalized_widths,
+            unnormalized_heights,
+            unnormalized_derivatives,
+            inverse=reverse,
+            tails="linear",
+            tail_bound=self.tail_bound,
+        )
+
+        x = torch.cat([x0, x1], 1) * x_mask
+        logdet = torch.sum(logabsdet * x_mask, [1, 2])
+        if not reverse:
+            return x, logdet
+        else:
+            return x

monotonic_align/__init__.py

@@ -0,0 +1,19 @@
+import numpy as np
+import torch
+from .monotonic_align.core import maximum_path_c
+
+
+def maximum_path(neg_cent, mask):
+    """Cython optimized version.
+    neg_cent: [b, t_t, t_s]
+    mask: [b, t_t, t_s]
+    """
+    device = neg_cent.device
+    dtype = neg_cent.dtype
+    neg_cent = neg_cent.data.cpu().numpy().astype(np.float32)
+    path = np.zeros(neg_cent.shape, dtype=np.int32)
+
+    t_t_max = mask.sum(1)[:, 0].data.cpu().numpy().astype(np.int32)
+    t_s_max = mask.sum(2)[:, 0].data.cpu().numpy().astype(np.int32)
+    maximum_path_c(path, neg_cent, t_t_max, t_s_max)
+    return torch.from_numpy(path).to(device=device, dtype=dtype)

monotonic_align/core.pyx

@@ -0,0 +1,42 @@
+cimport cython
+from cython.parallel import prange
+
+
+@cython.boundscheck(False)
+@cython.wraparound(False)
+cdef void maximum_path_each(int[:,::1] path, float[:,::1] value, int t_y, int t_x, float max_neg_val=-1e9) nogil:
+  cdef int x
+  cdef int y
+  cdef float v_prev
+  cdef float v_cur
+  cdef float tmp
+  cdef int index = t_x - 1
+
+  for y in range(t_y):
+    for x in range(max(0, t_x + y - t_y), min(t_x, y + 1)):
+      if x == y:
+        v_cur = max_neg_val
+      else:
+        v_cur = value[y-1, x]
+      if x == 0:
+        if y == 0:
+          v_prev = 0.
+        else:
+          v_prev = max_neg_val
+      else:
+        v_prev = value[y-1, x-1]
+      value[y, x] += max(v_prev, v_cur)
+
+  for y in range(t_y - 1, -1, -1):
+    path[y, index] = 1
+    if index != 0 and (index == y or value[y-1, index] < value[y-1, index-1]):
+      index = index - 1
+
+
+@cython.boundscheck(False)
+@cython.wraparound(False)
+cpdef void maximum_path_c(int[:,:,::1] paths, float[:,:,::1] values, int[::1] t_ys, int[::1] t_xs) nogil:
+  cdef int b = paths.shape[0]
+  cdef int i
+  for i in prange(b, nogil=True):
+    maximum_path_each(paths[i], values[i], t_ys[i], t_xs[i])

monotonic_align/monotonic_align/.gitkeep

@@ -0,0 +1 @@
+

monotonic_align/setup.py

@@ -0,0 +1,9 @@
+from distutils.core import setup
+from Cython.Build import cythonize
+import numpy
+
+setup(
+    name="monotonic_align",
+    ext_modules=cythonize("core.pyx"),
+    include_dirs=[numpy.get_include()],
+)

preprocess.py

@@ -0,0 +1,76 @@
+import os
+import argparse
+import json
+import sys
+sys.setrecursionlimit(500000)  # Fix the error message of RecursionError: maximum recursion depth exceeded while calling a Python object.  You can change the number as you want.
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--add_auxiliary_data", type=bool, default="False", help="Whether to add extra data as fine-tuning helper")
+    parser.add_argument("--languages", default="C")
+    args = parser.parse_args()
+    if args.languages == "CJE":
+        langs = ["[ZH]", "[JA]", "[EN]"]
+    elif args.languages == "CJ":
+        langs = ["[ZH]", "[JA]"]
+    elif args.languages == "C":
+        langs = ["[ZH]"]
+    new_annos = []
+    # Source 1: transcribed short audios
+    if os.path.exists("./filelists/short_character_anno.list"):
+        with open("./filelists/short_character_anno.list", 'r', encoding='utf-8') as f:
+            short_character_anno = f.readlines()
+            new_annos += short_character_anno
+
+    # Get all speaker names
+    speakers = []
+    for line in new_annos:
+        path, speaker, text = line.split("|")
+        if speaker not in speakers:
+            speakers.append(speaker)
+    assert (len(speakers) != 0), "No audio file found. Please check your uploaded file structure."
+    if True:
+        # Do not add extra helper data
+        # STEP 1: modify config file
+        with open("./configs/finetune_speaker.json", 'r', encoding='utf-8') as f:
+            hps = json.load(f)
+
+        # assign ids to new speakers
+        speaker2id = {}
+        for i, speaker in enumerate(speakers):
+            speaker2id[speaker] = i
+        # modify n_speakers
+        hps['data']["n_speakers"] = len(speakers)
+        # overwrite speaker names
+        hps['speakers'] = speaker2id
+        hps['train']['log_interval'] = 10
+        hps['train']['eval_interval'] = 100
+        hps['train']['batch_size'] = 16
+        hps['data']['training_files'] = "final_annotation_train.txt"
+        hps['data']['validation_files'] = "final_annotation_val.txt"
+        # save modified config
+        with open("./configs/modified_finetune_speaker.json", 'w', encoding='utf-8') as f:
+            json.dump(hps, f, indent=2)
+
+        # STEP 2: clean annotations, replace speaker names with assigned speaker IDs
+        import text
+
+        cleaned_new_annos = []
+        for i, line in enumerate(new_annos):
+            path, speaker, txt = line.split("|")
+            if len(txt) > 150:
+                continue
+            cleaned_text = text._clean_text(txt, hps['data']['text_cleaners']).replace("[ZH]", "")
+            cleaned_text += "\n" if not cleaned_text.endswith("\n") else ""
+            cleaned_new_annos.append(path + "|" + cleaned_text)
+
+        final_annos = cleaned_new_annos
+        # save annotation file
+        with open("./filelists/final_annotation_train.txt", 'w', encoding='utf-8') as f:
+            for line in final_annos:
+                f.write(line)
+        # save annotation file for validation
+        with open("./filelists/final_annotation_val.txt", 'w', encoding='utf-8') as f:
+            for line in cleaned_new_annos:
+                f.write(line)
+        print("finished")

requirements.txt

@@ -0,0 +1,29 @@
+librosa==0.10.1
+onnx==1.14.1
+onnxruntime==1.15.1
+matplotlib
+numpy
+numba
+phonemizer
+scipy
+tensorboard
+torch
+torchaudio
+torchvision
+Unidecode
+amfm_decompy
+jieba
+transformers
+pypinyin
+cn2an
+gradio
+av
+mecab-python3
+loguru
+unidic-lite
+cmudict
+fugashi
+num2words
+Cython==0.29.21
+openai-whisper
+protobuf==3.20.*

short_audio_transcribe.py

@@ -0,0 +1,122 @@
+import whisper
+import os
+import json
+import torchaudio
+import argparse
+import torch
+
+lang2token = {
+            'zh': "[ZH]",
+            'ja': "[JA]",
+            "en": "[EN]",
+        }
+def transcribe_one(audio_path):
+    # load audio and pad/trim it to fit 30 seconds
+    audio = whisper.load_audio(audio_path)
+    audio = whisper.pad_or_trim(audio)
+
+    # make log-Mel spectrogram and move to the same device as the model
+    mel = whisper.log_mel_spectrogram(audio).to(model.device)
+
+    # detect the spoken language
+    _, probs = model.detect_language(mel)
+    print(f"Detected language: {max(probs, key=probs.get)}")
+    lang = max(probs, key=probs.get)
+    # decode the audio
+    options = whisper.DecodingOptions(beam_size=5)
+    result = whisper.decode(model, mel, options)
+
+    # print the recognized text
+    print(result.text)
+    return lang, result.text
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--languages", default="CJE")
+    parser.add_argument("--whisper_size", default="medium")
+    args = parser.parse_args()
+    if args.languages == "CJE":
+        lang2token = {
+            'zh': "[ZH]",
+            'ja': "[JA]",
+            "en": "[EN]",
+        }
+    elif args.languages == "CJ":
+        lang2token = {
+            'zh': "[ZH]",
+            'ja': "[JA]",
+        }
+    elif args.languages == "C":
+        lang2token = {
+            'zh': "[ZH]",
+        }
+    assert (torch.cuda.is_available()), "Please enable GPU in order to run Whisper!"
+    model = whisper.load_model(args.whisper_size)
+    parent_dir = "./custom_character_voice/"
+    speaker_names = list(os.walk(parent_dir))[0][1]
+    speaker_annos = []
+    total_files = sum([len(files) for r, d, files in os.walk(parent_dir)])
+    # resample audios
+    # 2023/4/21: Get the target sampling rate
+    with open("./configs/config.json", 'r', encoding='utf-8') as f:
+        hps = json.load(f)
+    target_sr = hps['data']['sampling_rate']
+    processed_files = 0
+    for speaker in speaker_names:
+        for i, wavfile in enumerate(list(os.walk(parent_dir + speaker))[0][2]):
+            # try to load file as audio
+            if wavfile.startswith("processed_"):
+                continue
+            try:
+                wav, sr = torchaudio.load(parent_dir + speaker + "/" + wavfile, frame_offset=0, num_frames=-1, normalize=True,
+                                          channels_first=True)
+                wav = wav.mean(dim=0).unsqueeze(0)
+                if sr != target_sr:
+                    wav = torchaudio.transforms.Resample(orig_freq=sr, new_freq=target_sr)(wav)
+                if wav.shape[1] / sr > 20:
+                    print(f"{wavfile} too long, ignoring\n")
+                save_path = parent_dir + speaker + "/" + f"processed_{i}.wav"
+                torchaudio.save(save_path, wav, target_sr, channels_first=True)
+                # transcribe text
+                lang, text = transcribe_one(save_path)
+                if lang not in list(lang2token.keys()):
+                    print(f"{lang} not supported, ignoring\n")
+                    continue
+                text = text + "\n"#
+                #text = lang2token[lang] + text + lang2token[lang] + "\n"
+                speaker_annos.append(save_path + "|" + "0" + "|" + text)
+                
+                processed_files += 1
+                print(f"Processed: {processed_files}/{total_files}")
+            except:
+                continue
+
+    # # clean annotation
+    # import argparse
+    # import text
+    # from utils import load_filepaths_and_text
+    # for i, line in enumerate(speaker_annos):
+    #     path, sid, txt = line.split("|")
+    #     cleaned_text = text._clean_text(txt, ["cjke_cleaners2"])
+    #     cleaned_text += "\n" if not cleaned_text.endswith("\n") else ""
+    #     speaker_annos[i] = path + "|" + sid + "|" + cleaned_text
+    # write into annotation
+    if len(speaker_annos) == 0:
+        print("Warning: no short audios found, this IS expected if you have only uploaded long audios, videos or video links.")
+        print("this IS NOT expected if you have uploaded a zip file of short audios. Please check your file structure or make sure your audio language is supported.")
+    with open("./filelists/short_character_anno.list", 'w', encoding='utf-8') as f:
+        for line in speaker_annos:
+            f.write(line)
+
+    # import json
+    # # generate new config
+    # with open("./configs/finetune_speaker.json", 'r', encoding='utf-8') as f:
+    #     hps = json.load(f)
+    # # modify n_speakers
+    # hps['data']["n_speakers"] = 1000 + len(speaker2id)
+    # # add speaker names
+    # for speaker in speaker_names:
+    #     hps['speakers'][speaker] = speaker2id[speaker]
+    # # save modified config
+    # with open("./configs/modified_finetune_speaker.json", 'w', encoding='utf-8') as f:
+    #     json.dump(hps, f, indent=2)
+    # print("finished")

text/LICENSE

@@ -0,0 +1,19 @@
+Copyright (c) 2017 Keith Ito
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.

text/__init__.py

@@ -0,0 +1,64 @@
+""" from https://github.com/keithito/tacotron """
+from text import cleaners
+from text.symbols import symbols
+
+
+# Mappings from symbol to numeric ID and vice versa:
+_symbol_to_id = {s: i for i, s in enumerate(symbols)}
+_id_to_symbol = {i: s for i, s in enumerate(symbols)}
+
+
+def text_to_sequence(text, cleaner_names):
+    """Converts a string of text to a sequence of IDs corresponding to the symbols in the text.
+    Args:
+      text: string to convert to a sequence
+      cleaner_names: names of the cleaner functions to run the text through
+    Returns:
+      List of integers corresponding to the symbols in the text
+    """
+    sequence = []
+
+    clean_text = _clean_text(text, cleaner_names)
+    for symbol in clean_text:
+        if symbol in _symbol_to_id.keys():
+            symbol_id = _symbol_to_id[symbol]
+            sequence += [symbol_id]
+        else:
+            continue
+    return sequence
+
+
+def cleaned_text_to_sequence(cleaned_text):
+    """Converts a string of text to a sequence of IDs corresponding to the symbols in the text.
+    Args:
+      text: string to convert to a sequence
+    Returns:
+      List of integers corresponding to the symbols in the text
+    """
+    sequence = []
+    
+    for symbol in cleaned_text:
+        if symbol in _symbol_to_id.keys():
+            symbol_id = _symbol_to_id[symbol]
+            sequence += [symbol_id]
+        else:
+            continue
+    return sequence
+
+
+def sequence_to_text(sequence):
+    """Converts a sequence of IDs back to a string"""
+    result = ""
+    for symbol_id in sequence:
+        s = _id_to_symbol[symbol_id]
+        result += s
+    return result
+
+
+def _clean_text(text, cleaner_names):
+    for name in cleaner_names:
+        cleaner = getattr(cleaners, name)
+        if not cleaner:
+            raise Exception("Unknown cleaner: %s" % name)
+        text = cleaner(text)
+    return text

text/cleaners.py

@@ -0,0 +1,13 @@
+import re
+
+from text.mandarin import number_to_chinese, chinese_to_bopomofo, latin_to_bopomofo, chinese_to_romaji, chinese_to_lazy_ipa, chinese_to_ipa, chinese_to_ipa2
+
+
+def chinese_cleaners(text):
+    '''Pipeline for Chinese text'''
+    text = text.replace("[ZH]", "")
+    text = number_to_chinese(text)
+    text = chinese_to_bopomofo(text)
+    text = latin_to_bopomofo(text)
+    text = re.sub(r'([ˉˊˇˋ˙])$', r'\1。', text)
+    return text

text/mandarin.py

@@ -0,0 +1,326 @@
+import os
+import sys
+import re
+from pypinyin import lazy_pinyin, BOPOMOFO
+import jieba
+import cn2an
+import logging
+
+
+# List of (Latin alphabet, bopomofo) pairs:
+_latin_to_bopomofo = [(re.compile('%s' % x[0], re.IGNORECASE), x[1]) for x in [
+    ('a', 'ㄟˉ'),
+    ('b', 'ㄅㄧˋ'),
+    ('c', 'ㄙㄧˉ'),
+    ('d', 'ㄉㄧˋ'),
+    ('e', 'ㄧˋ'),
+    ('f', 'ㄝˊㄈㄨˋ'),
+    ('g', 'ㄐㄧˋ'),
+    ('h', 'ㄝˇㄑㄩˋ'),
+    ('i', 'ㄞˋ'),
+    ('j', 'ㄐㄟˋ'),
+    ('k', 'ㄎㄟˋ'),
+    ('l', 'ㄝˊㄛˋ'),
+    ('m', 'ㄝˊㄇㄨˋ'),
+    ('n', 'ㄣˉ'),
+    ('o', 'ㄡˉ'),
+    ('p', 'ㄆㄧˉ'),
+    ('q', 'ㄎㄧㄡˉ'),
+    ('r', 'ㄚˋ'),
+    ('s', 'ㄝˊㄙˋ'),
+    ('t', 'ㄊㄧˋ'),
+    ('u', 'ㄧㄡˉ'),
+    ('v', 'ㄨㄧˉ'),
+    ('w', 'ㄉㄚˋㄅㄨˋㄌㄧㄡˋ'),
+    ('x', 'ㄝˉㄎㄨˋㄙˋ'),
+    ('y', 'ㄨㄞˋ'),
+    ('z', 'ㄗㄟˋ')
+]]
+
+# List of (bopomofo, romaji) pairs:
+_bopomofo_to_romaji = [(re.compile('%s' % x[0]), x[1]) for x in [
+    ('ㄅㄛ', 'p⁼wo'),
+    ('ㄆㄛ', 'pʰwo'),
+    ('ㄇㄛ', 'mwo'),
+    ('ㄈㄛ', 'fwo'),
+    ('ㄅ', 'p⁼'),
+    ('ㄆ', 'pʰ'),
+    ('ㄇ', 'm'),
+    ('ㄈ', 'f'),
+    ('ㄉ', 't⁼'),
+    ('ㄊ', 'tʰ'),
+    ('ㄋ', 'n'),
+    ('ㄌ', 'l'),
+    ('ㄍ', 'k⁼'),
+    ('ㄎ', 'kʰ'),
+    ('ㄏ', 'h'),
+    ('ㄐ', 'ʧ⁼'),
+    ('ㄑ', 'ʧʰ'),
+    ('ㄒ', 'ʃ'),
+    ('ㄓ', 'ʦ`⁼'),
+    ('ㄔ', 'ʦ`ʰ'),
+    ('ㄕ', 's`'),
+    ('ㄖ', 'ɹ`'),
+    ('ㄗ', 'ʦ⁼'),
+    ('ㄘ', 'ʦʰ'),
+    ('ㄙ', 's'),
+    ('ㄚ', 'a'),
+    ('ㄛ', 'o'),
+    ('ㄜ', 'ə'),
+    ('ㄝ', 'e'),
+    ('ㄞ', 'ai'),
+    ('ㄟ', 'ei'),
+    ('ㄠ', 'au'),
+    ('ㄡ', 'ou'),
+    ('ㄧㄢ', 'yeNN'),
+    ('ㄢ', 'aNN'),
+    ('ㄧㄣ', 'iNN'),
+    ('ㄣ', 'əNN'),
+    ('ㄤ', 'aNg'),
+    ('ㄧㄥ', 'iNg'),
+    ('ㄨㄥ', 'uNg'),
+    ('ㄩㄥ', 'yuNg'),
+    ('ㄥ', 'əNg'),
+    ('ㄦ', 'əɻ'),
+    ('ㄧ', 'i'),
+    ('ㄨ', 'u'),
+    ('ㄩ', 'ɥ'),
+    ('ˉ', '→'),
+    ('ˊ', '↑'),
+    ('ˇ', '↓↑'),
+    ('ˋ', '↓'),
+    ('˙', ''),
+    ('，', ','),
+    ('。', '.'),
+    ('！', '!'),
+    ('？', '?'),
+    ('—', '-')
+]]
+
+# List of (romaji, ipa) pairs:
+_romaji_to_ipa = [(re.compile('%s' % x[0], re.IGNORECASE), x[1]) for x in [
+    ('ʃy', 'ʃ'),
+    ('ʧʰy', 'ʧʰ'),
+    ('ʧ⁼y', 'ʧ⁼'),
+    ('NN', 'n'),
+    ('Ng', 'ŋ'),
+    ('y', 'j'),
+    ('h', 'x')
+]]
+
+# List of (bopomofo, ipa) pairs:
+_bopomofo_to_ipa = [(re.compile('%s' % x[0]), x[1]) for x in [
+    ('ㄅㄛ', 'p⁼wo'),
+    ('ㄆㄛ', 'pʰwo'),
+    ('ㄇㄛ', 'mwo'),
+    ('ㄈㄛ', 'fwo'),
+    ('ㄅ', 'p⁼'),
+    ('ㄆ', 'pʰ'),
+    ('ㄇ', 'm'),
+    ('ㄈ', 'f'),
+    ('ㄉ', 't⁼'),
+    ('ㄊ', 'tʰ'),
+    ('ㄋ', 'n'),
+    ('ㄌ', 'l'),
+    ('ㄍ', 'k⁼'),
+    ('ㄎ', 'kʰ'),
+    ('ㄏ', 'x'),
+    ('ㄐ', 'tʃ⁼'),
+    ('ㄑ', 'tʃʰ'),
+    ('ㄒ', 'ʃ'),
+    ('ㄓ', 'ts`⁼'),
+    ('ㄔ', 'ts`ʰ'),
+    ('ㄕ', 's`'),
+    ('ㄖ', 'ɹ`'),
+    ('ㄗ', 'ts⁼'),
+    ('ㄘ', 'tsʰ'),
+    ('ㄙ', 's'),
+    ('ㄚ', 'a'),
+    ('ㄛ', 'o'),
+    ('ㄜ', 'ə'),
+    ('ㄝ', 'ɛ'),
+    ('ㄞ', 'aɪ'),
+    ('ㄟ', 'eɪ'),
+    ('ㄠ', 'ɑʊ'),
+    ('ㄡ', 'oʊ'),
+    ('ㄧㄢ', 'jɛn'),
+    ('ㄩㄢ', 'ɥæn'),
+    ('ㄢ', 'an'),
+    ('ㄧㄣ', 'in'),
+    ('ㄩㄣ', 'ɥn'),
+    ('ㄣ', 'ən'),
+    ('ㄤ', 'ɑŋ'),
+    ('ㄧㄥ', 'iŋ'),
+    ('ㄨㄥ', 'ʊŋ'),
+    ('ㄩㄥ', 'jʊŋ'),
+    ('ㄥ', 'əŋ'),
+    ('ㄦ', 'əɻ'),
+    ('ㄧ', 'i'),
+    ('ㄨ', 'u'),
+    ('ㄩ', 'ɥ'),
+    ('ˉ', '→'),
+    ('ˊ', '↑'),
+    ('ˇ', '↓↑'),
+    ('ˋ', '↓'),
+    ('˙', ''),
+    ('，', ','),
+    ('。', '.'),
+    ('！', '!'),
+    ('？', '?'),
+    ('—', '-')
+]]
+
+# List of (bopomofo, ipa2) pairs:
+_bopomofo_to_ipa2 = [(re.compile('%s' % x[0]), x[1]) for x in [
+    ('ㄅㄛ', 'pwo'),
+    ('ㄆㄛ', 'pʰwo'),
+    ('ㄇㄛ', 'mwo'),
+    ('ㄈㄛ', 'fwo'),
+    ('ㄅ', 'p'),
+    ('ㄆ', 'pʰ'),
+    ('ㄇ', 'm'),
+    ('ㄈ', 'f'),
+    ('ㄉ', 't'),
+    ('ㄊ', 'tʰ'),
+    ('ㄋ', 'n'),
+    ('ㄌ', 'l'),
+    ('ㄍ', 'k'),
+    ('ㄎ', 'kʰ'),
+    ('ㄏ', 'h'),
+    ('ㄐ', 'tɕ'),
+    ('ㄑ', 'tɕʰ'),
+    ('ㄒ', 'ɕ'),
+    ('ㄓ', 'tʂ'),
+    ('ㄔ', 'tʂʰ'),
+    ('ㄕ', 'ʂ'),
+    ('ㄖ', 'ɻ'),
+    ('ㄗ', 'ts'),
+    ('ㄘ', 'tsʰ'),
+    ('ㄙ', 's'),
+    ('ㄚ', 'a'),
+    ('ㄛ', 'o'),
+    ('ㄜ', 'ɤ'),
+    ('ㄝ', 'ɛ'),
+    ('ㄞ', 'aɪ'),
+    ('ㄟ', 'eɪ'),
+    ('ㄠ', 'ɑʊ'),
+    ('ㄡ', 'oʊ'),
+    ('ㄧㄢ', 'jɛn'),
+    ('ㄩㄢ', 'yæn'),
+    ('ㄢ', 'an'),
+    ('ㄧㄣ', 'in'),
+    ('ㄩㄣ', 'yn'),
+    ('ㄣ', 'ən'),
+    ('ㄤ', 'ɑŋ'),
+    ('ㄧㄥ', 'iŋ'),
+    ('ㄨㄥ', 'ʊŋ'),
+    ('ㄩㄥ', 'jʊŋ'),
+    ('ㄥ', 'ɤŋ'),
+    ('ㄦ', 'əɻ'),
+    ('ㄧ', 'i'),
+    ('ㄨ', 'u'),
+    ('ㄩ', 'y'),
+    ('ˉ', '˥'),
+    ('ˊ', '˧˥'),
+    ('ˇ', '˨˩˦'),
+    ('ˋ', '˥˩'),
+    ('˙', ''),
+    ('，', ','),
+    ('。', '.'),
+    ('！', '!'),
+    ('？', '?'),
+    ('—', '-')
+]]
+
+
+def number_to_chinese(text):
+    numbers = re.findall(r'\d+(?:\.?\d+)?', text)
+    for number in numbers:
+        text = text.replace(number, cn2an.an2cn(number), 1)
+    return text
+
+
+def chinese_to_bopomofo(text):
+    text = text.replace('、', '，').replace('；', '，').replace('：', '，')
+    words = jieba.lcut(text, cut_all=False)
+    text = ''
+    for word in words:
+        bopomofos = lazy_pinyin(word, BOPOMOFO)
+        if not re.search('[\u4e00-\u9fff]', word):
+            text += word
+            continue
+        for i in range(len(bopomofos)):
+            bopomofos[i] = re.sub(r'([\u3105-\u3129])$', r'\1ˉ', bopomofos[i])
+        if text != '':
+            text += ' '
+        text += ''.join(bopomofos)
+    return text
+
+
+def latin_to_bopomofo(text):
+    for regex, replacement in _latin_to_bopomofo:
+        text = re.sub(regex, replacement, text)
+    return text
+
+
+def bopomofo_to_romaji(text):
+    for regex, replacement in _bopomofo_to_romaji:
+        text = re.sub(regex, replacement, text)
+    return text
+
+
+def bopomofo_to_ipa(text):
+    for regex, replacement in _bopomofo_to_ipa:
+        text = re.sub(regex, replacement, text)
+    return text
+
+
+def bopomofo_to_ipa2(text):
+    for regex, replacement in _bopomofo_to_ipa2:
+        text = re.sub(regex, replacement, text)
+    return text
+
+
+def chinese_to_romaji(text):
+    text = number_to_chinese(text)
+    text = chinese_to_bopomofo(text)
+    text = latin_to_bopomofo(text)
+    text = bopomofo_to_romaji(text)
+    text = re.sub('i([aoe])', r'y\1', text)
+    text = re.sub('u([aoəe])', r'w\1', text)
+    text = re.sub('([ʦsɹ]`[⁼ʰ]?)([→↓↑ ]+|$)',
+                  r'\1ɹ`\2', text).replace('ɻ', 'ɹ`')
+    text = re.sub('([ʦs][⁼ʰ]?)([→↓↑ ]+|$)', r'\1ɹ\2', text)
+    return text
+
+
+def chinese_to_lazy_ipa(text):
+    text = chinese_to_romaji(text)
+    for regex, replacement in _romaji_to_ipa:
+        text = re.sub(regex, replacement, text)
+    return text
+
+
+def chinese_to_ipa(text):
+    text = number_to_chinese(text)
+    text = chinese_to_bopomofo(text)
+    text = latin_to_bopomofo(text)
+    text = bopomofo_to_ipa(text)
+    text = re.sub('i([aoe])', r'j\1', text)
+    text = re.sub('u([aoəe])', r'w\1', text)
+    text = re.sub('([sɹ]`[⁼ʰ]?)([→↓↑ ]+|$)',
+                  r'\1ɹ`\2', text).replace('ɻ', 'ɹ`')
+    text = re.sub('([s][⁼ʰ]?)([→↓↑ ]+|$)', r'\1ɹ\2', text)
+    return text
+
+
+def chinese_to_ipa2(text):
+    text = number_to_chinese(text)
+    text = chinese_to_bopomofo(text)
+    text = latin_to_bopomofo(text)
+    text = bopomofo_to_ipa2(text)
+    text = re.sub(r'i([aoe])', r'j\1', text)
+    text = re.sub(r'u([aoəe])', r'w\1', text)
+    text = re.sub(r'([ʂɹ]ʰ?)([˩˨˧˦˥ ]+|$)', r'\1ʅ\2', text)
+    text = re.sub(r'(sʰ?)([˩˨˧˦˥ ]+|$)', r'\1ɿ\2', text)
+    return text

text/symbols.py

@@ -0,0 +1,14 @@
+'''
+Defines the set of symbols used in text input to the model.
+'''
+
+# chinese_cleaners
+_pad        = '_'
+_punctuation = '，,。！？—…'
+_letters = 'ㄅㄆㄇㄈㄉㄊㄋㄌㄍㄎㄏㄐㄑㄒㄓㄔㄕㄖㄗㄘㄙㄚㄛㄜㄝㄞㄟㄠㄡㄢㄣㄤㄥㄦㄧㄨㄩˉˊˇˋ˙ '
+
+# Export all symbols:
+symbols = [_pad] + list(_punctuation) + list(_letters)
+
+# Special symbol ids
+SPACE_ID = symbols.index(" ")

train.py

@@ -0,0 +1,585 @@
+import os
+import json
+import argparse
+import itertools
+import math
+import torch
+from torch import nn, optim
+from torch.nn import functional as F
+from torch.utils.data import DataLoader
+from torch.utils.tensorboard import SummaryWriter
+
+# from tensorboardX import SummaryWriter
+import torch.multiprocessing as mp
+import torch.distributed as dist
+from torch.nn.parallel import DistributedDataParallel as DDP
+from torch.cuda.amp import autocast, GradScaler
+import tqdm
+
+import commons
+import utils
+from data_utils import TextAudioLoader, TextAudioCollate, DistributedBucketSampler
+from models import (
+    SynthesizerTrn,
+    MultiPeriodDiscriminator,
+    DurationDiscriminator,
+    AVAILABLE_FLOW_TYPES,
+)
+from losses import generator_loss, discriminator_loss, feature_loss, kl_loss
+from mel_processing import mel_spectrogram_torch, spec_to_mel_torch
+from text.symbols import symbols
+
+
+torch.backends.cudnn.benchmark = True
+global_step = 0
+
+
+def main():
+    """Assume Single Node Multi GPUs Training Only"""
+    assert torch.cuda.is_available(), "CPU training is not allowed."
+
+    n_gpus = torch.cuda.device_count()
+    os.environ["MASTER_ADDR"] = "localhost"
+    os.environ["MASTER_PORT"] = "6060"
+
+    hps = utils.get_hparams()
+    mp.spawn(
+        run,
+        nprocs=n_gpus,
+        args=(
+            n_gpus,
+            hps,
+        ),
+    )
+
+
+def run(rank, n_gpus, hps):
+    global global_step
+    if rank == 0:
+        logger = utils.get_logger(hps.model_dir)
+        logger.info(hps)
+        utils.check_git_hash(hps.model_dir)
+        writer = SummaryWriter(log_dir=hps.model_dir)
+        writer_eval = SummaryWriter(log_dir=os.path.join(hps.model_dir, "eval"))
+
+    dist.init_process_group(
+        backend="nccl", init_method="env://", world_size=n_gpus, rank=rank
+    )
+    torch.manual_seed(hps.train.seed)
+    torch.cuda.set_device(rank)
+
+    if (
+        "use_mel_posterior_encoder" in hps.model.keys()
+        and hps.model.use_mel_posterior_encoder == True
+    ):
+        print("Using mel posterior encoder for VITS2")
+        posterior_channels = 80  # vits2
+        hps.data.use_mel_posterior_encoder = True
+    else:
+        print("Using lin posterior encoder for VITS1")
+        posterior_channels = hps.data.filter_length // 2 + 1
+        hps.data.use_mel_posterior_encoder = False
+
+    train_dataset = TextAudioLoader(hps.data.training_files, hps.data)
+    train_sampler = DistributedBucketSampler(
+        train_dataset,
+        hps.train.batch_size,
+        [32, 300, 400, 500, 600, 700, 800, 900, 1000],
+        num_replicas=n_gpus,
+        rank=rank,
+        shuffle=True,
+    )
+
+    collate_fn = TextAudioCollate()
+    train_loader = DataLoader(
+        train_dataset,
+        num_workers=8,
+        shuffle=False,
+        pin_memory=True,
+        collate_fn=collate_fn,
+        batch_sampler=train_sampler,
+    )
+    if rank == 0:
+        eval_dataset = TextAudioLoader(hps.data.validation_files, hps.data)
+        eval_loader = DataLoader(
+            eval_dataset,
+            num_workers=8,
+            shuffle=False,
+            batch_size=hps.train.batch_size,
+            pin_memory=True,
+            drop_last=False,
+            collate_fn=collate_fn,
+        )
+    # some of these flags are not being used in the code and directly set in hps json file.
+    # they are kept here for reference and prototyping.
+
+    if (
+        "use_transformer_flows" in hps.model.keys()
+        and hps.model.use_transformer_flows == True
+    ):
+        use_transformer_flows = True
+        transformer_flow_type = hps.model.transformer_flow_type
+        print(f"Using transformer flows {transformer_flow_type} for VITS2")
+        assert (
+            transformer_flow_type in AVAILABLE_FLOW_TYPES
+        ), f"transformer_flow_type must be one of {AVAILABLE_FLOW_TYPES}"
+    else:
+        print("Using normal flows for VITS1")
+        use_transformer_flows = False
+
+    if (
+        "use_spk_conditioned_encoder" in hps.model.keys()
+        and hps.model.use_spk_conditioned_encoder == True
+    ):
+        if hps.data.n_speakers == 0:
+            print("Warning: use_spk_conditioned_encoder is True but n_speakers is 0")
+        print(
+            "Setting use_spk_conditioned_encoder to False as model is a single speaker model"
+        )
+        use_spk_conditioned_encoder = False
+    else:
+        print("Using normal encoder for VITS1")
+        use_spk_conditioned_encoder = False
+
+    if (
+        "use_noise_scaled_mas" in hps.model.keys()
+        and hps.model.use_noise_scaled_mas == True
+    ):
+        print("Using noise scaled MAS for VITS2")
+        use_noise_scaled_mas = True
+        mas_noise_scale_initial = 0.01
+        noise_scale_delta = 2e-6
+    else:
+        print("Using normal MAS for VITS1")
+        use_noise_scaled_mas = False
+        mas_noise_scale_initial = 0.0
+        noise_scale_delta = 0.0
+
+    if (
+        "use_duration_discriminator" in hps.model.keys()
+        and hps.model.use_duration_discriminator == True
+    ):
+        print("Using duration discriminator for VITS2")
+        use_duration_discriminator = True
+        net_dur_disc = DurationDiscriminator(
+            hps.model.hidden_channels,
+            hps.model.hidden_channels,
+            3,
+            0.1,
+            gin_channels=hps.model.gin_channels if hps.data.n_speakers != 0 else 0,
+        ).cuda(rank)
+    else:
+        print("NOT using any duration discriminator like VITS1")
+        net_dur_disc = None
+        use_duration_discriminator = False
+
+    net_g = SynthesizerTrn(
+        len(symbols),
+        posterior_channels,
+        hps.train.segment_size // hps.data.hop_length,
+        mas_noise_scale_initial=mas_noise_scale_initial,
+        noise_scale_delta=noise_scale_delta,
+        **hps.model,
+    ).cuda(rank)
+    net_d = MultiPeriodDiscriminator(hps.model.use_spectral_norm).cuda(rank)
+    optim_g = torch.optim.AdamW(
+        net_g.parameters(),
+        hps.train.learning_rate,
+        betas=hps.train.betas,
+        eps=hps.train.eps,
+    )
+    optim_d = torch.optim.AdamW(
+        net_d.parameters(),
+        hps.train.learning_rate,
+        betas=hps.train.betas,
+        eps=hps.train.eps,
+    )
+    if net_dur_disc is not None:
+        optim_dur_disc = torch.optim.AdamW(
+            net_dur_disc.parameters(),
+            hps.train.learning_rate,
+            betas=hps.train.betas,
+            eps=hps.train.eps,
+        )
+    else:
+        optim_dur_disc = None
+
+    net_g = DDP(net_g, device_ids=[rank], find_unused_parameters=True)
+    net_d = DDP(net_d, device_ids=[rank], find_unused_parameters=True)
+    if net_dur_disc is not None:
+        net_dur_disc = DDP(net_dur_disc, device_ids=[rank], find_unused_parameters=True)
+
+    try:
+        _, _, _, epoch_str = utils.load_checkpoint(
+            utils.latest_checkpoint_path(hps.model_dir, "G_*.pth"), net_g, optim_g
+        )
+        _, _, _, epoch_str = utils.load_checkpoint(
+            utils.latest_checkpoint_path(hps.model_dir, "D_*.pth"), net_d, optim_d
+        )
+        if net_dur_disc is not None:
+            _, _, _, epoch_str = utils.load_checkpoint(
+                utils.latest_checkpoint_path(hps.model_dir, "DUR_*.pth"),
+                net_dur_disc,
+                optim_dur_disc,
+            )
+        global_step = (epoch_str - 1) * len(train_loader)
+    except:
+        epoch_str = 1
+        global_step = 0
+
+    scheduler_g = torch.optim.lr_scheduler.ExponentialLR(
+        optim_g, gamma=hps.train.lr_decay, last_epoch=epoch_str - 2
+    )
+    scheduler_d = torch.optim.lr_scheduler.ExponentialLR(
+        optim_d, gamma=hps.train.lr_decay, last_epoch=epoch_str - 2
+    )
+    if net_dur_disc is not None:
+        scheduler_dur_disc = torch.optim.lr_scheduler.ExponentialLR(
+            optim_dur_disc, gamma=hps.train.lr_decay, last_epoch=epoch_str - 2
+        )
+    else:
+        scheduler_dur_disc = None
+
+    scaler = GradScaler(enabled=hps.train.fp16_run)
+
+    for epoch in range(epoch_str, hps.train.epochs + 1):
+        if rank == 0:
+            train_and_evaluate(
+                rank,
+                epoch,
+                hps,
+                [net_g, net_d, net_dur_disc],
+                [optim_g, optim_d, optim_dur_disc],
+                [scheduler_g, scheduler_d, scheduler_dur_disc],
+                scaler,
+                [train_loader, eval_loader],
+                logger,
+                [writer, writer_eval],
+            )
+        else:
+            train_and_evaluate(
+                rank,
+                epoch,
+                hps,
+                [net_g, net_d, net_dur_disc],
+                [optim_g, optim_d, optim_dur_disc],
+                [scheduler_g, scheduler_d, scheduler_dur_disc],
+                scaler,
+                [train_loader, None],
+                None,
+                None,
+            )
+        scheduler_g.step()
+        scheduler_d.step()
+        if net_dur_disc is not None:
+            scheduler_dur_disc.step()
+
+
+def train_and_evaluate(
+    rank, epoch, hps, nets, optims, schedulers, scaler, loaders, logger, writers
+):
+    net_g, net_d, net_dur_disc = nets
+    optim_g, optim_d, optim_dur_disc = optims
+    scheduler_g, scheduler_d, scheduler_dur_disc = schedulers
+    train_loader, eval_loader = loaders
+    if writers is not None:
+        writer, writer_eval = writers
+
+    train_loader.batch_sampler.set_epoch(epoch)
+    global global_step
+
+    net_g.train()
+    net_d.train()
+    if net_dur_disc is not None:
+        net_dur_disc.train()
+
+    if rank == 0:
+        loader = tqdm.tqdm(train_loader, desc="Loading train data")
+    else:
+        loader = train_loader
+    for batch_idx, (x, x_lengths, spec, spec_lengths, y, y_lengths) in enumerate(
+        loader
+    ):
+        if net_g.module.use_noise_scaled_mas:
+            current_mas_noise_scale = (
+                net_g.module.mas_noise_scale_initial
+                - net_g.module.noise_scale_delta * global_step
+            )
+            net_g.module.current_mas_noise_scale = max(current_mas_noise_scale, 0.0)
+        x, x_lengths = x.cuda(rank, non_blocking=True), x_lengths.cuda(
+            rank, non_blocking=True
+        )
+        spec, spec_lengths = spec.cuda(rank, non_blocking=True), spec_lengths.cuda(
+            rank, non_blocking=True
+        )
+        y, y_lengths = y.cuda(rank, non_blocking=True), y_lengths.cuda(
+            rank, non_blocking=True
+        )
+
+        with autocast(enabled=hps.train.fp16_run):
+            (
+                y_hat,
+                l_length,
+                attn,
+                ids_slice,
+                x_mask,
+                z_mask,
+                (z, z_p, m_p, logs_p, m_q, logs_q),
+                (hidden_x, logw, logw_),
+            ) = net_g(x, x_lengths, spec, spec_lengths)
+
+            if (
+                hps.model.use_mel_posterior_encoder
+                or hps.data.use_mel_posterior_encoder
+            ):
+                mel = spec
+            else:
+                mel = spec_to_mel_torch(
+                    spec,
+                    hps.data.filter_length,
+                    hps.data.n_mel_channels,
+                    hps.data.sampling_rate,
+                    hps.data.mel_fmin,
+                    hps.data.mel_fmax,
+                )
+            y_mel = commons.slice_segments(
+                mel, ids_slice, hps.train.segment_size // hps.data.hop_length
+            )
+            y_hat_mel = mel_spectrogram_torch(
+                y_hat.squeeze(1),
+                hps.data.filter_length,
+                hps.data.n_mel_channels,
+                hps.data.sampling_rate,
+                hps.data.hop_length,
+                hps.data.win_length,
+                hps.data.mel_fmin,
+                hps.data.mel_fmax,
+            )
+
+            y = commons.slice_segments(
+                y, ids_slice * hps.data.hop_length, hps.train.segment_size
+            )  # slice
+
+            # Discriminator
+            y_d_hat_r, y_d_hat_g, _, _ = net_d(y, y_hat.detach())
+            with autocast(enabled=False):
+                loss_disc, losses_disc_r, losses_disc_g = discriminator_loss(
+                    y_d_hat_r, y_d_hat_g
+                )
+                loss_disc_all = loss_disc
+
+            # Duration Discriminator
+            if net_dur_disc is not None:
+                y_dur_hat_r, y_dur_hat_g = net_dur_disc(
+                    hidden_x.detach(), x_mask.detach(), logw_.detach(), logw.detach()
+                )
+                with autocast(enabled=False):
+                    # TODO: I think need to mean using the mask, but for now, just mean all
+                    (
+                        loss_dur_disc,
+                        losses_dur_disc_r,
+                        losses_dur_disc_g,
+                    ) = discriminator_loss(y_dur_hat_r, y_dur_hat_g)
+                    loss_dur_disc_all = loss_dur_disc
+                optim_dur_disc.zero_grad()
+                scaler.scale(loss_dur_disc_all).backward()
+                scaler.unscale_(optim_dur_disc)
+                grad_norm_dur_disc = commons.clip_grad_value_(
+                    net_dur_disc.parameters(), None
+                )
+                scaler.step(optim_dur_disc)
+
+        optim_d.zero_grad()
+        scaler.scale(loss_disc_all).backward()
+        scaler.unscale_(optim_d)
+        grad_norm_d = commons.clip_grad_value_(net_d.parameters(), None)
+        scaler.step(optim_d)
+
+        with autocast(enabled=hps.train.fp16_run):
+            # Generator
+            y_d_hat_r, y_d_hat_g, fmap_r, fmap_g = net_d(y, y_hat)
+            if net_dur_disc is not None:
+                y_dur_hat_r, y_dur_hat_g = net_dur_disc(hidden_x, x_mask, logw_, logw)
+            with autocast(enabled=False):
+                loss_dur = torch.sum(l_length.float())
+                loss_mel = F.l1_loss(y_mel, y_hat_mel) * hps.train.c_mel
+                loss_kl = kl_loss(z_p, logs_q, m_p, logs_p, z_mask) * hps.train.c_kl
+
+                loss_fm = feature_loss(fmap_r, fmap_g)
+                loss_gen, losses_gen = generator_loss(y_d_hat_g)
+                loss_gen_all = loss_gen + loss_fm + loss_mel + loss_dur + loss_kl
+                if net_dur_disc is not None:
+                    loss_dur_gen, losses_dur_gen = generator_loss(y_dur_hat_g)
+                    loss_gen_all += loss_dur_gen
+
+        optim_g.zero_grad()
+        scaler.scale(loss_gen_all).backward()
+        scaler.unscale_(optim_g)
+        grad_norm_g = commons.clip_grad_value_(net_g.parameters(), None)
+        scaler.step(optim_g)
+        scaler.update()
+
+        if rank == 0:
+            if global_step % hps.train.log_interval == 0:
+                lr = optim_g.param_groups[0]["lr"]
+                losses = [loss_disc, loss_gen, loss_fm, loss_mel, loss_dur, loss_kl]
+                logger.info(
+                    "Train Epoch: {} [{:.0f}%]".format(
+                        epoch, 100.0 * batch_idx / len(train_loader)
+                    )
+                )
+                logger.info([x.item() for x in losses] + [global_step, lr])
+
+                scalar_dict = {
+                    "loss/g/total": loss_gen_all,
+                    "loss/d/total": loss_disc_all,
+                    "learning_rate": lr,
+                    "grad_norm_d": grad_norm_d,
+                    "grad_norm_g": grad_norm_g,
+                }
+                if net_dur_disc is not None:
+                    scalar_dict.update(
+                        {
+                            "loss/dur_disc/total": loss_dur_disc_all,
+                            "grad_norm_dur_disc": grad_norm_dur_disc,
+                        }
+                    )
+                scalar_dict.update(
+                    {
+                        "loss/g/fm": loss_fm,
+                        "loss/g/mel": loss_mel,
+                        "loss/g/dur": loss_dur,
+                        "loss/g/kl": loss_kl,
+                    }
+                )
+
+                scalar_dict.update(
+                    {"loss/g/{}".format(i): v for i, v in enumerate(losses_gen)}
+                )
+                scalar_dict.update(
+                    {"loss/d_r/{}".format(i): v for i, v in enumerate(losses_disc_r)}
+                )
+                scalar_dict.update(
+                    {"loss/d_g/{}".format(i): v for i, v in enumerate(losses_disc_g)}
+                )
+
+                # if net_dur_disc is not None:
+                #   scalar_dict.update({"loss/dur_disc_r" : f"{losses_dur_disc_r}"})
+                #   scalar_dict.update({"loss/dur_disc_g" : f"{losses_dur_disc_g}"})
+                #   scalar_dict.update({"loss/dur_gen" : f"{loss_dur_gen}"})
+
+                image_dict = {
+                    "slice/mel_org": utils.plot_spectrogram_to_numpy(
+                        y_mel[0].data.cpu().numpy()
+                    ),
+                    "slice/mel_gen": utils.plot_spectrogram_to_numpy(
+                        y_hat_mel[0].data.cpu().numpy()
+                    ),
+                    "all/mel": utils.plot_spectrogram_to_numpy(
+                        mel[0].data.cpu().numpy()
+                    ),
+                    "all/attn": utils.plot_alignment_to_numpy(
+                        attn[0, 0].data.cpu().numpy()
+                    ),
+                }
+                utils.summarize(
+                    writer=writer,
+                    global_step=global_step,
+                    images=image_dict,
+                    scalars=scalar_dict,
+                )
+
+            if global_step % hps.train.eval_interval == 0:
+                evaluate(hps, net_g, eval_loader, writer_eval)
+                utils.save_checkpoint(
+                    net_g,
+                    optim_g,
+                    hps.train.learning_rate,
+                    epoch,
+                    os.path.join(hps.model_dir, "G_{}.pth".format(global_step)),
+                )
+                utils.save_checkpoint(
+                    net_d,
+                    optim_d,
+                    hps.train.learning_rate,
+                    epoch,
+                    os.path.join(hps.model_dir, "D_{}.pth".format(global_step)),
+                )
+                if net_dur_disc is not None:
+                    utils.save_checkpoint(
+                        net_dur_disc,
+                        optim_dur_disc,
+                        hps.train.learning_rate,
+                        epoch,
+                        os.path.join(hps.model_dir, "DUR_{}.pth".format(global_step)),
+                    )
+        global_step += 1
+
+    if rank == 0:
+        logger.info("====> Epoch: {}".format(epoch))
+
+
+def evaluate(hps, generator, eval_loader, writer_eval):
+    generator.eval()
+    with torch.no_grad():
+        for batch_idx, (x, x_lengths, spec, spec_lengths, y, y_lengths) in enumerate(
+            eval_loader
+        ):
+            x, x_lengths = x.cuda(0), x_lengths.cuda(0)
+            spec, spec_lengths = spec.cuda(0), spec_lengths.cuda(0)
+            y, y_lengths = y.cuda(0), y_lengths.cuda(0)
+
+            # remove else
+            x = x[:1]
+            x_lengths = x_lengths[:1]
+            spec = spec[:1]
+            spec_lengths = spec_lengths[:1]
+            y = y[:1]
+            y_lengths = y_lengths[:1]
+            break
+        y_hat, attn, mask, *_ = generator.module.infer(x, x_lengths, max_len=1000)
+        y_hat_lengths = mask.sum([1, 2]).long() * hps.data.hop_length
+
+        if hps.model.use_mel_posterior_encoder or hps.data.use_mel_posterior_encoder:
+            mel = spec
+        else:
+            mel = spec_to_mel_torch(
+                spec,
+                hps.data.filter_length,
+                hps.data.n_mel_channels,
+                hps.data.sampling_rate,
+                hps.data.mel_fmin,
+                hps.data.mel_fmax,
+            )
+        y_hat_mel = mel_spectrogram_torch(
+            y_hat.squeeze(1).float(),
+            hps.data.filter_length,
+            hps.data.n_mel_channels,
+            hps.data.sampling_rate,
+            hps.data.hop_length,
+            hps.data.win_length,
+            hps.data.mel_fmin,
+            hps.data.mel_fmax,
+        )
+    image_dict = {
+        "gen/mel": utils.plot_spectrogram_to_numpy(y_hat_mel[0].cpu().numpy())
+    }
+    audio_dict = {"gen/audio": y_hat[0, :, : y_hat_lengths[0]]}
+    if global_step == 0:
+        image_dict.update(
+            {"gt/mel": utils.plot_spectrogram_to_numpy(mel[0].cpu().numpy())}
+        )
+        audio_dict.update({"gt/audio": y[0, :, : y_lengths[0]]})
+
+    utils.summarize(
+        writer=writer_eval,
+        global_step=global_step,
+        images=image_dict,
+        audios=audio_dict,
+        audio_sampling_rate=hps.data.sampling_rate,
+    )
+    generator.train()
+
+
+if __name__ == "__main__":
+    main()

train_ms.py

@@ -0,0 +1,604 @@
+import os
+import json
+import argparse
+import itertools
+import math
+import torch
+from torch import nn, optim
+from torch.nn import functional as F
+from torch.utils.data import DataLoader
+from torch.utils.tensorboard import SummaryWriter
+
+# from tensorboardX import SummaryWriter
+import torch.multiprocessing as mp
+import torch.distributed as dist
+from torch.nn.parallel import DistributedDataParallel as DDP
+from torch.cuda.amp import autocast, GradScaler
+import tqdm
+
+import commons
+import utils
+from data_utils import (
+    TextAudioSpeakerLoader,
+    TextAudioSpeakerCollate,
+    DistributedBucketSampler,
+)
+from models import (
+    SynthesizerTrn,
+    MultiPeriodDiscriminator,
+    DurationDiscriminator,
+    AVAILABLE_FLOW_TYPES,
+)
+from losses import generator_loss, discriminator_loss, feature_loss, kl_loss
+from mel_processing import mel_spectrogram_torch, spec_to_mel_torch
+from text.symbols import symbols
+
+
+torch.backends.cudnn.benchmark = True
+global_step = 0
+
+
+def main():
+    """Assume Single Node Multi GPUs Training Only"""
+    assert torch.cuda.is_available(), "CPU training is not allowed."
+
+    n_gpus = torch.cuda.device_count()
+    os.environ["MASTER_ADDR"] = "localhost"
+    os.environ["MASTER_PORT"] = "6060"
+
+    hps = utils.get_hparams()
+    mp.spawn(
+        run,
+        nprocs=n_gpus,
+        args=(
+            n_gpus,
+            hps,
+        ),
+    )
+
+
+def run(rank, n_gpus, hps):
+    global global_step
+    if rank == 0:
+        logger = utils.get_logger(hps.model_dir)
+        logger.info(hps)
+        utils.check_git_hash(hps.model_dir)
+        writer = SummaryWriter(log_dir=hps.model_dir)
+        writer_eval = SummaryWriter(log_dir=os.path.join(hps.model_dir, "eval"))
+
+    dist.init_process_group(
+        backend="nccl", init_method="env://", world_size=n_gpus, rank=rank
+    )
+    torch.manual_seed(hps.train.seed)
+    torch.cuda.set_device(rank)
+
+    if (
+        "use_mel_posterior_encoder" in hps.model.keys()
+        and hps.model.use_mel_posterior_encoder == True
+    ):
+        print("Using mel posterior encoder for VITS2")
+        posterior_channels = 80  # vits2
+        hps.data.use_mel_posterior_encoder = True
+    else:
+        print("Using lin posterior encoder for VITS1")
+        posterior_channels = hps.data.filter_length // 2 + 1
+        hps.data.use_mel_posterior_encoder = False
+
+    train_dataset = TextAudioSpeakerLoader(hps.data.training_files, hps.data)
+    train_sampler = DistributedBucketSampler(
+        train_dataset,
+        hps.train.batch_size,
+        [32, 300, 400, 500, 600, 700, 800, 900, 1000],
+        num_replicas=n_gpus,
+        rank=rank,
+        shuffle=True,
+    )
+    collate_fn = TextAudioSpeakerCollate()
+    train_loader = DataLoader(
+        train_dataset,
+        num_workers=8,
+        shuffle=False,
+        pin_memory=True,
+        collate_fn=collate_fn,
+        batch_sampler=train_sampler,
+    )
+    if rank == 0:
+        eval_dataset = TextAudioSpeakerLoader(hps.data.validation_files, hps.data)
+        eval_loader = DataLoader(
+            eval_dataset,
+            num_workers=8,
+            shuffle=False,
+            batch_size=hps.train.batch_size,
+            pin_memory=True,
+            drop_last=False,
+            collate_fn=collate_fn,
+        )
+    # some of these flags are not being used in the code and directly set in hps json file.
+    # they are kept here for reference and prototyping.
+    if (
+        "use_transformer_flows" in hps.model.keys()
+        and hps.model.use_transformer_flows == True
+    ):
+        use_transformer_flows = True
+        transformer_flow_type = hps.model.transformer_flow_type
+        print(f"Using transformer flows {transformer_flow_type} for VITS2")
+        assert (
+            transformer_flow_type in AVAILABLE_FLOW_TYPES
+        ), f"transformer_flow_type must be one of {AVAILABLE_FLOW_TYPES}"
+    else:
+        print("Using normal flows for VITS1")
+        use_transformer_flows = False
+
+    if (
+        "use_spk_conditioned_encoder" in hps.model.keys()
+        and hps.model.use_spk_conditioned_encoder == True
+    ):
+        if hps.data.n_speakers == 0:
+            raise ValueError(
+                "n_speakers must be > 0 when using spk conditioned encoder to train multi-speaker model"
+            )
+        use_spk_conditioned_encoder = True
+    else:
+        print("Using normal encoder for VITS1")
+        use_spk_conditioned_encoder = False
+
+    if (
+        "use_noise_scaled_mas" in hps.model.keys()
+        and hps.model.use_noise_scaled_mas == True
+    ):
+        print("Using noise scaled MAS for VITS2")
+        use_noise_scaled_mas = True
+        mas_noise_scale_initial = 0.01
+        noise_scale_delta = 2e-6
+    else:
+        print("Using normal MAS for VITS1")
+        use_noise_scaled_mas = False
+        mas_noise_scale_initial = 0.0
+        noise_scale_delta = 0.0
+
+    if (
+        "use_duration_discriminator" in hps.model.keys()
+        and hps.model.use_duration_discriminator == True
+    ):
+        print("Using duration discriminator for VITS2")
+        use_duration_discriminator = True
+        net_dur_disc = DurationDiscriminator(
+            hps.model.hidden_channels,
+            hps.model.hidden_channels,
+            3,
+            0.1,
+            gin_channels=hps.model.gin_channels if hps.data.n_speakers != 0 else 0,
+        ).cuda(rank)
+    else:
+        print("NOT using any duration discriminator like VITS1")
+        net_dur_disc = None
+        use_duration_discriminator = False
+
+    net_g = SynthesizerTrn(
+        len(symbols),
+        posterior_channels,
+        hps.train.segment_size // hps.data.hop_length,
+        n_speakers=hps.data.n_speakers,
+        mas_noise_scale_initial=mas_noise_scale_initial,
+        noise_scale_delta=noise_scale_delta,
+        **hps.model,
+    ).cuda(rank)
+    net_d = MultiPeriodDiscriminator(hps.model.use_spectral_norm).cuda(rank)
+    optim_g = torch.optim.AdamW(
+        net_g.parameters(),
+        hps.train.learning_rate,
+        betas=hps.train.betas,
+        eps=hps.train.eps,
+    )
+    optim_d = torch.optim.AdamW(
+        net_d.parameters(),
+        hps.train.learning_rate,
+        betas=hps.train.betas,
+        eps=hps.train.eps,
+    )
+    if net_dur_disc is not None:
+        optim_dur_disc = torch.optim.AdamW(
+            net_dur_disc.parameters(),
+            hps.train.learning_rate,
+            betas=hps.train.betas,
+            eps=hps.train.eps,
+        )
+    else:
+        optim_dur_disc = None
+
+    net_g = DDP(net_g, device_ids=[rank], find_unused_parameters=True)
+    net_d = DDP(net_d, device_ids=[rank], find_unused_parameters=True)
+    if net_dur_disc is not None:
+        net_dur_disc = DDP(net_dur_disc, device_ids=[rank], find_unused_parameters=True)
+
+    try:
+        _, _, _, epoch_str = utils.load_checkpoint(
+            utils.latest_checkpoint_path(hps.model_dir, "G_*.pth"), net_g, optim_g
+        )
+        _, _, _, epoch_str = utils.load_checkpoint(
+            utils.latest_checkpoint_path(hps.model_dir, "D_*.pth"), net_d, optim_d
+        )
+        if net_dur_disc is not None:
+            _, _, _, epoch_str = utils.load_checkpoint(
+                utils.latest_checkpoint_path(hps.model_dir, "DUR_*.pth"),
+                net_dur_disc,
+                optim_dur_disc,
+            )
+        global_step = (epoch_str - 1) * len(train_loader)
+    except:
+        epoch_str = 1
+        global_step = 0
+
+    scheduler_g = torch.optim.lr_scheduler.ExponentialLR(
+        optim_g, gamma=hps.train.lr_decay, last_epoch=epoch_str - 2
+    )
+    scheduler_d = torch.optim.lr_scheduler.ExponentialLR(
+        optim_d, gamma=hps.train.lr_decay, last_epoch=epoch_str - 2
+    )
+    if net_dur_disc is not None:
+        scheduler_dur_disc = torch.optim.lr_scheduler.ExponentialLR(
+            optim_dur_disc, gamma=hps.train.lr_decay, last_epoch=epoch_str - 2
+        )
+    else:
+        scheduler_dur_disc = None
+
+    scaler = GradScaler(enabled=hps.train.fp16_run)
+
+    for epoch in range(epoch_str, hps.train.epochs + 1):
+        if rank == 0:
+            train_and_evaluate(
+                rank,
+                epoch,
+                hps,
+                [net_g, net_d, net_dur_disc],
+                [optim_g, optim_d, optim_dur_disc],
+                [scheduler_g, scheduler_d, scheduler_dur_disc],
+                scaler,
+                [train_loader, eval_loader],
+                logger,
+                [writer, writer_eval],
+            )
+        else:
+            train_and_evaluate(
+                rank,
+                epoch,
+                hps,
+                [net_g, net_d, net_dur_disc],
+                [optim_g, optim_d, optim_dur_disc],
+                [scheduler_g, scheduler_d, scheduler_dur_disc],
+                scaler,
+                [train_loader, None],
+                None,
+                None,
+            )
+        scheduler_g.step()
+        scheduler_d.step()
+        if net_dur_disc is not None:
+            scheduler_dur_disc.step()
+
+
+def train_and_evaluate(
+    rank, epoch, hps, nets, optims, schedulers, scaler, loaders, logger, writers
+):
+    net_g, net_d, net_dur_disc = nets
+    optim_g, optim_d, optim_dur_disc = optims
+    scheduler_g, scheduler_d, scheduler_dur_disc = schedulers
+    train_loader, eval_loader = loaders
+    if writers is not None:
+        writer, writer_eval = writers
+
+    train_loader.batch_sampler.set_epoch(epoch)
+    global global_step
+
+    net_g.train()
+    net_d.train()
+    if net_dur_disc is not None:
+        net_dur_disc.train()
+
+    if rank == 0:
+        loader = tqdm.tqdm(train_loader, desc="Loading train data")
+    else:
+        loader = train_loader
+    for batch_idx, (
+        x,
+        x_lengths,
+        spec,
+        spec_lengths,
+        y,
+        y_lengths,
+        speakers,
+    ) in enumerate(loader):
+        if net_g.module.use_noise_scaled_mas:
+            current_mas_noise_scale = (
+                net_g.module.mas_noise_scale_initial
+                - net_g.module.noise_scale_delta * global_step
+            )
+            net_g.module.current_mas_noise_scale = max(current_mas_noise_scale, 0.0)
+        x, x_lengths = x.cuda(rank, non_blocking=True), x_lengths.cuda(
+            rank, non_blocking=True
+        )
+        spec, spec_lengths = spec.cuda(rank, non_blocking=True), spec_lengths.cuda(
+            rank, non_blocking=True
+        )
+        y, y_lengths = y.cuda(rank, non_blocking=True), y_lengths.cuda(
+            rank, non_blocking=True
+        )
+        speakers = speakers.cuda(rank, non_blocking=True)
+
+        with autocast(enabled=hps.train.fp16_run):
+            (
+                y_hat,
+                l_length,
+                attn,
+                ids_slice,
+                x_mask,
+                z_mask,
+                (z, z_p, m_p, logs_p, m_q, logs_q),
+                (hidden_x, logw, logw_),
+            ) = net_g(x, x_lengths, spec, spec_lengths, speakers)
+
+            if (
+                hps.model.use_mel_posterior_encoder
+                or hps.data.use_mel_posterior_encoder
+            ):
+                mel = spec
+            else:
+                mel = spec_to_mel_torch(
+                    spec,
+                    hps.data.filter_length,
+                    hps.data.n_mel_channels,
+                    hps.data.sampling_rate,
+                    hps.data.mel_fmin,
+                    hps.data.mel_fmax,
+                )
+            y_mel = commons.slice_segments(
+                mel, ids_slice, hps.train.segment_size // hps.data.hop_length
+            )
+            y_hat_mel = mel_spectrogram_torch(
+                y_hat.squeeze(1),
+                hps.data.filter_length,
+                hps.data.n_mel_channels,
+                hps.data.sampling_rate,
+                hps.data.hop_length,
+                hps.data.win_length,
+                hps.data.mel_fmin,
+                hps.data.mel_fmax,
+            )
+
+            y = commons.slice_segments(
+                y, ids_slice * hps.data.hop_length, hps.train.segment_size
+            )  # slice
+
+            # Discriminator
+            y_d_hat_r, y_d_hat_g, _, _ = net_d(y, y_hat.detach())
+            with autocast(enabled=False):
+                loss_disc, losses_disc_r, losses_disc_g = discriminator_loss(
+                    y_d_hat_r, y_d_hat_g
+                )
+                loss_disc_all = loss_disc
+
+            # Duration Discriminator
+            if net_dur_disc is not None:
+                y_dur_hat_r, y_dur_hat_g = net_dur_disc(
+                    hidden_x.detach(), x_mask.detach(), logw_.detach(), logw.detach()
+                )
+                with autocast(enabled=False):
+                    # TODO: I think need to mean using the mask, but for now, just mean all
+                    (
+                        loss_dur_disc,
+                        losses_dur_disc_r,
+                        losses_dur_disc_g,
+                    ) = discriminator_loss(y_dur_hat_r, y_dur_hat_g)
+                    loss_dur_disc_all = loss_dur_disc
+                optim_dur_disc.zero_grad()
+                scaler.scale(loss_dur_disc_all).backward()
+                scaler.unscale_(optim_dur_disc)
+                grad_norm_dur_disc = commons.clip_grad_value_(
+                    net_dur_disc.parameters(), None
+                )
+                scaler.step(optim_dur_disc)
+
+        optim_d.zero_grad()
+        scaler.scale(loss_disc_all).backward()
+        scaler.unscale_(optim_d)
+        grad_norm_d = commons.clip_grad_value_(net_d.parameters(), None)
+        scaler.step(optim_d)
+
+        with autocast(enabled=hps.train.fp16_run):
+            # Generator
+            y_d_hat_r, y_d_hat_g, fmap_r, fmap_g = net_d(y, y_hat)
+            if net_dur_disc is not None:
+                y_dur_hat_r, y_dur_hat_g = net_dur_disc(hidden_x, x_mask, logw_, logw)
+            with autocast(enabled=False):
+                loss_dur = torch.sum(l_length.float())
+                loss_mel = F.l1_loss(y_mel, y_hat_mel) * hps.train.c_mel
+                loss_kl = kl_loss(z_p, logs_q, m_p, logs_p, z_mask) * hps.train.c_kl
+
+                loss_fm = feature_loss(fmap_r, fmap_g)
+                loss_gen, losses_gen = generator_loss(y_d_hat_g)
+                loss_gen_all = loss_gen + loss_fm + loss_mel + loss_dur + loss_kl
+                if net_dur_disc is not None:
+                    loss_dur_gen, losses_dur_gen = generator_loss(y_dur_hat_g)
+                    loss_gen_all += loss_dur_gen
+
+        optim_g.zero_grad()
+        scaler.scale(loss_gen_all).backward()
+        scaler.unscale_(optim_g)
+        grad_norm_g = commons.clip_grad_value_(net_g.parameters(), None)
+        scaler.step(optim_g)
+        scaler.update()
+
+        if rank == 0:
+            if global_step % hps.train.log_interval == 0:
+                lr = optim_g.param_groups[0]["lr"]
+                losses = [loss_disc, loss_gen, loss_fm, loss_mel, loss_dur, loss_kl]
+                logger.info(
+                    "Train Epoch: {} [{:.0f}%]".format(
+                        epoch, 100.0 * batch_idx / len(train_loader)
+                    )
+                )
+                logger.info([x.item() for x in losses] + [global_step, lr])
+
+                scalar_dict = {
+                    "loss/g/total": loss_gen_all,
+                    "loss/d/total": loss_disc_all,
+                    "learning_rate": lr,
+                    "grad_norm_d": grad_norm_d,
+                    "grad_norm_g": grad_norm_g,
+                }
+                if net_dur_disc is not None:
+                    scalar_dict.update(
+                        {
+                            "loss/dur_disc/total": loss_dur_disc_all,
+                            "grad_norm_dur_disc": grad_norm_dur_disc,
+                        }
+                    )
+                scalar_dict.update(
+                    {
+                        "loss/g/fm": loss_fm,
+                        "loss/g/mel": loss_mel,
+                        "loss/g/dur": loss_dur,
+                        "loss/g/kl": loss_kl,
+                    }
+                )
+
+                scalar_dict.update(
+                    {"loss/g/{}".format(i): v for i, v in enumerate(losses_gen)}
+                )
+                scalar_dict.update(
+                    {"loss/d_r/{}".format(i): v for i, v in enumerate(losses_disc_r)}
+                )
+                scalar_dict.update(
+                    {"loss/d_g/{}".format(i): v for i, v in enumerate(losses_disc_g)}
+                )
+
+                # if net_dur_disc is not None:
+                #   scalar_dict.update({"loss/dur_disc_r" : f"{losses_dur_disc_r}"})
+                #   scalar_dict.update({"loss/dur_disc_g" : f"{losses_dur_disc_g}"})
+                #   scalar_dict.update({"loss/dur_gen" : f"{loss_dur_gen}"})
+
+                image_dict = {
+                    "slice/mel_org": utils.plot_spectrogram_to_numpy(
+                        y_mel[0].data.cpu().numpy()
+                    ),
+                    "slice/mel_gen": utils.plot_spectrogram_to_numpy(
+                        y_hat_mel[0].data.cpu().numpy()
+                    ),
+                    "all/mel": utils.plot_spectrogram_to_numpy(
+                        mel[0].data.cpu().numpy()
+                    ),
+                    "all/attn": utils.plot_alignment_to_numpy(
+                        attn[0, 0].data.cpu().numpy()
+                    ),
+                }
+                utils.summarize(
+                    writer=writer,
+                    global_step=global_step,
+                    images=image_dict,
+                    scalars=scalar_dict,
+                )
+
+            if global_step % hps.train.eval_interval == 0:
+                evaluate(hps, net_g, eval_loader, writer_eval)
+                utils.save_checkpoint(
+                    net_g,
+                    optim_g,
+                    hps.train.learning_rate,
+                    epoch,
+                    os.path.join(hps.model_dir, "G_{}.pth".format(global_step)),
+                )
+                utils.save_checkpoint(
+                    net_d,
+                    optim_d,
+                    hps.train.learning_rate,
+                    epoch,
+                    os.path.join(hps.model_dir, "D_{}.pth".format(global_step)),
+                )
+                if net_dur_disc is not None:
+                    utils.save_checkpoint(
+                        net_dur_disc,
+                        optim_dur_disc,
+                        hps.train.learning_rate,
+                        epoch,
+                        os.path.join(hps.model_dir, "DUR_{}.pth".format(global_step)),
+                    )
+        global_step += 1
+
+    if rank == 0:
+        logger.info("====> Epoch: {}".format(epoch))
+
+
+def evaluate(hps, generator, eval_loader, writer_eval):
+    generator.eval()
+    with torch.no_grad():
+        for batch_idx, (
+            x,
+            x_lengths,
+            spec,
+            spec_lengths,
+            y,
+            y_lengths,
+            speakers,
+        ) in enumerate(eval_loader):
+            x, x_lengths = x.cuda(0), x_lengths.cuda(0)
+            spec, spec_lengths = spec.cuda(0), spec_lengths.cuda(0)
+            y, y_lengths = y.cuda(0), y_lengths.cuda(0)
+            speakers = speakers.cuda(0)
+
+            # remove else
+            x = x[:1]
+            x_lengths = x_lengths[:1]
+            spec = spec[:1]
+            spec_lengths = spec_lengths[:1]
+            y = y[:1]
+            y_lengths = y_lengths[:1]
+            speakers = speakers[:1]
+            break
+        y_hat, attn, mask, *_ = generator.module.infer(
+            x, x_lengths, speakers, max_len=1000
+        )
+        y_hat_lengths = mask.sum([1, 2]).long() * hps.data.hop_length
+
+        if hps.model.use_mel_posterior_encoder or hps.data.use_mel_posterior_encoder:
+            mel = spec
+        else:
+            mel = spec_to_mel_torch(
+                spec,
+                hps.data.filter_length,
+                hps.data.n_mel_channels,
+                hps.data.sampling_rate,
+                hps.data.mel_fmin,
+                hps.data.mel_fmax,
+            )
+        y_hat_mel = mel_spectrogram_torch(
+            y_hat.squeeze(1).float(),
+            hps.data.filter_length,
+            hps.data.n_mel_channels,
+            hps.data.sampling_rate,
+            hps.data.hop_length,
+            hps.data.win_length,
+            hps.data.mel_fmin,
+            hps.data.mel_fmax,
+        )
+    image_dict = {
+        "gen/mel": utils.plot_spectrogram_to_numpy(y_hat_mel[0].cpu().numpy())
+    }
+    audio_dict = {"gen/audio": y_hat[0, :, : y_hat_lengths[0]]}
+    if global_step == 0:
+        image_dict.update(
+            {"gt/mel": utils.plot_spectrogram_to_numpy(mel[0].cpu().numpy())}
+        )
+        audio_dict.update({"gt/audio": y[0, :, : y_lengths[0]]})
+
+    utils.summarize(
+        writer=writer_eval,
+        global_step=global_step,
+        images=image_dict,
+        audios=audio_dict,
+        audio_sampling_rate=hps.data.sampling_rate,
+    )
+    generator.train()
+
+
+if __name__ == "__main__":
+    main()

transforms.py

@@ -0,0 +1,209 @@
+import torch
+from torch.nn import functional as F
+
+import numpy as np
+
+
+DEFAULT_MIN_BIN_WIDTH = 1e-3
+DEFAULT_MIN_BIN_HEIGHT = 1e-3
+DEFAULT_MIN_DERIVATIVE = 1e-3
+
+
+def piecewise_rational_quadratic_transform(
+    inputs,
+    unnormalized_widths,
+    unnormalized_heights,
+    unnormalized_derivatives,
+    inverse=False,
+    tails=None,
+    tail_bound=1.0,
+    min_bin_width=DEFAULT_MIN_BIN_WIDTH,
+    min_bin_height=DEFAULT_MIN_BIN_HEIGHT,
+    min_derivative=DEFAULT_MIN_DERIVATIVE,
+):
+    if tails is None:
+        spline_fn = rational_quadratic_spline
+        spline_kwargs = {}
+    else:
+        spline_fn = unconstrained_rational_quadratic_spline
+        spline_kwargs = {"tails": tails, "tail_bound": tail_bound}
+
+    outputs, logabsdet = spline_fn(
+        inputs=inputs,
+        unnormalized_widths=unnormalized_widths,
+        unnormalized_heights=unnormalized_heights,
+        unnormalized_derivatives=unnormalized_derivatives,
+        inverse=inverse,
+        min_bin_width=min_bin_width,
+        min_bin_height=min_bin_height,
+        min_derivative=min_derivative,
+        **spline_kwargs
+    )
+    return outputs, logabsdet
+
+
+def searchsorted(bin_locations, inputs, eps=1e-6):
+    bin_locations[..., -1] += eps
+    return torch.sum(inputs[..., None] >= bin_locations, dim=-1) - 1
+
+
+def unconstrained_rational_quadratic_spline(
+    inputs,
+    unnormalized_widths,
+    unnormalized_heights,
+    unnormalized_derivatives,
+    inverse=False,
+    tails="linear",
+    tail_bound=1.0,
+    min_bin_width=DEFAULT_MIN_BIN_WIDTH,
+    min_bin_height=DEFAULT_MIN_BIN_HEIGHT,
+    min_derivative=DEFAULT_MIN_DERIVATIVE,
+):
+    inside_interval_mask = (inputs >= -tail_bound) & (inputs <= tail_bound)
+    outside_interval_mask = ~inside_interval_mask
+
+    outputs = torch.zeros_like(inputs)
+    logabsdet = torch.zeros_like(inputs)
+
+    if tails == "linear":
+        unnormalized_derivatives = F.pad(unnormalized_derivatives, pad=(1, 1))
+        constant = np.log(np.exp(1 - min_derivative) - 1)
+        unnormalized_derivatives[..., 0] = constant
+        unnormalized_derivatives[..., -1] = constant
+
+        outputs[outside_interval_mask] = inputs[outside_interval_mask]
+        logabsdet[outside_interval_mask] = 0
+    else:
+        raise RuntimeError("{} tails are not implemented.".format(tails))
+
+    (
+        outputs[inside_interval_mask],
+        logabsdet[inside_interval_mask],
+    ) = rational_quadratic_spline(
+        inputs=inputs[inside_interval_mask],
+        unnormalized_widths=unnormalized_widths[inside_interval_mask, :],
+        unnormalized_heights=unnormalized_heights[inside_interval_mask, :],
+        unnormalized_derivatives=unnormalized_derivatives[inside_interval_mask, :],
+        inverse=inverse,
+        left=-tail_bound,
+        right=tail_bound,
+        bottom=-tail_bound,
+        top=tail_bound,
+        min_bin_width=min_bin_width,
+        min_bin_height=min_bin_height,
+        min_derivative=min_derivative,
+    )
+
+    return outputs, logabsdet
+
+
+def rational_quadratic_spline(
+    inputs,
+    unnormalized_widths,
+    unnormalized_heights,
+    unnormalized_derivatives,
+    inverse=False,
+    left=0.0,
+    right=1.0,
+    bottom=0.0,
+    top=1.0,
+    min_bin_width=DEFAULT_MIN_BIN_WIDTH,
+    min_bin_height=DEFAULT_MIN_BIN_HEIGHT,
+    min_derivative=DEFAULT_MIN_DERIVATIVE,
+):
+    if torch.min(inputs) < left or torch.max(inputs) > right:
+        raise ValueError("Input to a transform is not within its domain")
+
+    num_bins = unnormalized_widths.shape[-1]
+
+    if min_bin_width * num_bins > 1.0:
+        raise ValueError("Minimal bin width too large for the number of bins")
+    if min_bin_height * num_bins > 1.0:
+        raise ValueError("Minimal bin height too large for the number of bins")
+
+    widths = F.softmax(unnormalized_widths, dim=-1)
+    widths = min_bin_width + (1 - min_bin_width * num_bins) * widths
+    cumwidths = torch.cumsum(widths, dim=-1)
+    cumwidths = F.pad(cumwidths, pad=(1, 0), mode="constant", value=0.0)
+    cumwidths = (right - left) * cumwidths + left
+    cumwidths[..., 0] = left
+    cumwidths[..., -1] = right
+    widths = cumwidths[..., 1:] - cumwidths[..., :-1]
+
+    derivatives = min_derivative + F.softplus(unnormalized_derivatives)
+
+    heights = F.softmax(unnormalized_heights, dim=-1)
+    heights = min_bin_height + (1 - min_bin_height * num_bins) * heights
+    cumheights = torch.cumsum(heights, dim=-1)
+    cumheights = F.pad(cumheights, pad=(1, 0), mode="constant", value=0.0)
+    cumheights = (top - bottom) * cumheights + bottom
+    cumheights[..., 0] = bottom
+    cumheights[..., -1] = top
+    heights = cumheights[..., 1:] - cumheights[..., :-1]
+
+    if inverse:
+        bin_idx = searchsorted(cumheights, inputs)[..., None]
+    else:
+        bin_idx = searchsorted(cumwidths, inputs)[..., None]
+
+    input_cumwidths = cumwidths.gather(-1, bin_idx)[..., 0]
+    input_bin_widths = widths.gather(-1, bin_idx)[..., 0]
+
+    input_cumheights = cumheights.gather(-1, bin_idx)[..., 0]
+    delta = heights / widths
+    input_delta = delta.gather(-1, bin_idx)[..., 0]
+
+    input_derivatives = derivatives.gather(-1, bin_idx)[..., 0]
+    input_derivatives_plus_one = derivatives[..., 1:].gather(-1, bin_idx)[..., 0]
+
+    input_heights = heights.gather(-1, bin_idx)[..., 0]
+
+    if inverse:
+        a = (inputs - input_cumheights) * (
+            input_derivatives + input_derivatives_plus_one - 2 * input_delta
+        ) + input_heights * (input_delta - input_derivatives)
+        b = input_heights * input_derivatives - (inputs - input_cumheights) * (
+            input_derivatives + input_derivatives_plus_one - 2 * input_delta
+        )
+        c = -input_delta * (inputs - input_cumheights)
+
+        discriminant = b.pow(2) - 4 * a * c
+        assert (discriminant >= 0).all()
+
+        root = (2 * c) / (-b - torch.sqrt(discriminant))
+        outputs = root * input_bin_widths + input_cumwidths
+
+        theta_one_minus_theta = root * (1 - root)
+        denominator = input_delta + (
+            (input_derivatives + input_derivatives_plus_one - 2 * input_delta)
+            * theta_one_minus_theta
+        )
+        derivative_numerator = input_delta.pow(2) * (
+            input_derivatives_plus_one * root.pow(2)
+            + 2 * input_delta * theta_one_minus_theta
+            + input_derivatives * (1 - root).pow(2)
+        )
+        logabsdet = torch.log(derivative_numerator) - 2 * torch.log(denominator)
+
+        return outputs, -logabsdet
+    else:
+        theta = (inputs - input_cumwidths) / input_bin_widths
+        theta_one_minus_theta = theta * (1 - theta)
+
+        numerator = input_heights * (
+            input_delta * theta.pow(2) + input_derivatives * theta_one_minus_theta
+        )
+        denominator = input_delta + (
+            (input_derivatives + input_derivatives_plus_one - 2 * input_delta)
+            * theta_one_minus_theta
+        )
+        outputs = input_cumheights + numerator / denominator
+
+        derivative_numerator = input_delta.pow(2) * (
+            input_derivatives_plus_one * theta.pow(2)
+            + 2 * input_delta * theta_one_minus_theta
+            + input_derivatives * (1 - theta).pow(2)
+        )
+        logabsdet = torch.log(derivative_numerator) - 2 * torch.log(denominator)
+
+        return outputs, logabsdet

utils.py

@@ -0,0 +1,286 @@
+import os
+import glob
+import sys
+import argparse
+import logging
+import json
+import subprocess
+import numpy as np
+from scipy.io.wavfile import read
+import torch
+
+MATPLOTLIB_FLAG = False
+
+logging.basicConfig(stream=sys.stdout, level=logging.DEBUG)
+logger = logging
+
+
+def load_checkpoint(checkpoint_path, model, optimizer=None):
+    assert os.path.isfile(checkpoint_path)
+    checkpoint_dict = torch.load(checkpoint_path, map_location="cpu")
+    iteration = checkpoint_dict["iteration"]
+    learning_rate = checkpoint_dict["learning_rate"]
+    if optimizer is not None:
+        optimizer.load_state_dict(checkpoint_dict["optimizer"])
+    saved_state_dict = checkpoint_dict["model"]
+    if hasattr(model, "module"):
+        state_dict = model.module.state_dict()
+    else:
+        state_dict = model.state_dict()
+    new_state_dict = {}
+    for k, v in state_dict.items():
+        try:
+            new_state_dict[k] = saved_state_dict[k]
+        except:
+            logger.info("%s is not in the checkpoint" % k)
+            new_state_dict[k] = v
+    if hasattr(model, "module"):
+        model.module.load_state_dict(new_state_dict)
+    else:
+        model.load_state_dict(new_state_dict)
+    logger.info(
+        "Loaded checkpoint '{}' (iteration {})".format(checkpoint_path, iteration)
+    )
+    return model, optimizer, learning_rate, iteration
+
+
+def save_checkpoint(model, optimizer, learning_rate, iteration, checkpoint_path):
+    logger.info(
+        "Saving model and optimizer state at iteration {} to {}".format(
+            iteration, checkpoint_path
+        )
+    )
+    if hasattr(model, "module"):
+        state_dict = model.module.state_dict()
+    else:
+        state_dict = model.state_dict()
+    torch.save(
+        {
+            "model": state_dict,
+            "iteration": iteration,
+            "optimizer": optimizer.state_dict(),
+            "learning_rate": learning_rate,
+        },
+        checkpoint_path,
+    )
+
+
+def summarize(
+    writer,
+    global_step,
+    scalars={},
+    histograms={},
+    images={},
+    audios={},
+    audio_sampling_rate=22050,
+):
+    for k, v in scalars.items():
+        writer.add_scalar(k, v, global_step)
+    for k, v in histograms.items():
+        writer.add_histogram(k, v, global_step)
+    for k, v in images.items():
+        writer.add_image(k, v, global_step, dataformats="HWC")
+    for k, v in audios.items():
+        writer.add_audio(k, v, global_step, audio_sampling_rate)
+
+
+def latest_checkpoint_path(dir_path, regex="G_*.pth"):
+    f_list = glob.glob(os.path.join(dir_path, regex))
+    f_list.sort(key=lambda f: int("".join(filter(str.isdigit, f))))
+    x = f_list[-1]
+    print(x)
+    return x
+
+
+def plot_spectrogram_to_numpy(spectrogram):
+    global MATPLOTLIB_FLAG
+    if not MATPLOTLIB_FLAG:
+        import matplotlib
+
+        matplotlib.use("Agg")
+        MATPLOTLIB_FLAG = True
+        mpl_logger = logging.getLogger("matplotlib")
+        mpl_logger.setLevel(logging.WARNING)
+    import matplotlib.pylab as plt
+    import numpy as np
+
+    fig, ax = plt.subplots(figsize=(10, 2))
+    im = ax.imshow(spectrogram, aspect="auto", origin="lower", interpolation="none")
+    plt.colorbar(im, ax=ax)
+    plt.xlabel("Frames")
+    plt.ylabel("Channels")
+    plt.tight_layout()
+
+    fig.canvas.draw()
+    data = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep="")
+    data = data.reshape(fig.canvas.get_width_height()[::-1] + (3,))
+    plt.close()
+    return data
+
+
+def plot_alignment_to_numpy(alignment, info=None):
+    global MATPLOTLIB_FLAG
+    if not MATPLOTLIB_FLAG:
+        import matplotlib
+
+        matplotlib.use("Agg")
+        MATPLOTLIB_FLAG = True
+        mpl_logger = logging.getLogger("matplotlib")
+        mpl_logger.setLevel(logging.WARNING)
+    import matplotlib.pylab as plt
+    import numpy as np
+
+    fig, ax = plt.subplots(figsize=(6, 4))
+    im = ax.imshow(
+        alignment.transpose(), aspect="auto", origin="lower", interpolation="none"
+    )
+    fig.colorbar(im, ax=ax)
+    xlabel = "Decoder timestep"
+    if info is not None:
+        xlabel += "\n\n" + info
+    plt.xlabel(xlabel)
+    plt.ylabel("Encoder timestep")
+    plt.tight_layout()
+
+    fig.canvas.draw()
+    data = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep="")
+    data = data.reshape(fig.canvas.get_width_height()[::-1] + (3,))
+    plt.close()
+    return data
+
+
+def load_wav_to_torch(full_path):
+    sampling_rate, data = read(full_path)
+    return torch.FloatTensor(data.astype(np.float32)), sampling_rate
+
+
+def load_filepaths_and_text(filename, split="|"):
+    with open(filename, encoding="utf-8") as f:
+        filepaths_and_text = [line.strip().split(split) for line in f]
+    return filepaths_and_text
+
+
+def get_hparams(init=True):
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "-c",
+        "--config",
+        type=str,
+        default="./configs/base.json",
+        help="JSON file for configuration",
+    )
+    parser.add_argument("-m", "--model", type=str, required=True, help="Model name")
+
+    args = parser.parse_args()
+    model_dir = os.path.join("./logs", args.model)
+
+    if not os.path.exists(model_dir):
+        os.makedirs(model_dir)
+
+    config_path = args.config
+    config_save_path = os.path.join(model_dir, "config.json")
+    if init:
+        with open(config_path, "r") as f:
+            data = f.read()
+        with open(config_save_path, "w") as f:
+            f.write(data)
+    else:
+        with open(config_save_path, "r") as f:
+            data = f.read()
+    config = json.loads(data)
+
+    hparams = HParams(**config)
+    hparams.model_dir = model_dir
+    return hparams
+
+
+def get_hparams_from_dir(model_dir):
+    config_save_path = os.path.join(model_dir, "config.json")
+    with open(config_save_path, "r") as f:
+        data = f.read()
+    config = json.loads(data)
+
+    hparams = HParams(**config)
+    hparams.model_dir = model_dir
+    return hparams
+
+
+def get_hparams_from_file(config_path):
+    with open(config_path, "r") as f:
+        data = f.read()
+    config = json.loads(data)
+
+    hparams = HParams(**config)
+    return hparams
+
+
+def check_git_hash(model_dir):
+    source_dir = os.path.dirname(os.path.realpath(__file__))
+    if not os.path.exists(os.path.join(source_dir, ".git")):
+        logger.warn(
+            "{} is not a git repository, therefore hash value comparison will be ignored.".format(
+                source_dir
+            )
+        )
+        return
+
+    cur_hash = subprocess.getoutput("git rev-parse HEAD")
+
+    path = os.path.join(model_dir, "githash")
+    if os.path.exists(path):
+        saved_hash = open(path).read()
+        if saved_hash != cur_hash:
+            logger.warn(
+                "git hash values are different. {}(saved) != {}(current)".format(
+                    saved_hash[:8], cur_hash[:8]
+                )
+            )
+    else:
+        open(path, "w").write(cur_hash)
+
+
+def get_logger(model_dir, filename="train.log"):
+    global logger
+    logger = logging.getLogger(os.path.basename(model_dir))
+    logger.setLevel(logging.DEBUG)
+
+    formatter = logging.Formatter("%(asctime)s\t%(name)s\t%(levelname)s\t%(message)s")
+    if not os.path.exists(model_dir):
+        os.makedirs(model_dir)
+    h = logging.FileHandler(os.path.join(model_dir, filename))
+    h.setLevel(logging.DEBUG)
+    h.setFormatter(formatter)
+    logger.addHandler(h)
+    return logger
+
+
+class HParams:
+    def __init__(self, **kwargs):
+        for k, v in kwargs.items():
+            if type(v) == dict:
+                v = HParams(**v)
+            self[k] = v
+
+    def keys(self):
+        return self.__dict__.keys()
+
+    def items(self):
+        return self.__dict__.items()
+
+    def values(self):
+        return self.__dict__.values()
+
+    def __len__(self):
+        return len(self.__dict__)
+
+    def __getitem__(self, key):
+        return getattr(self, key)
+
+    def __setitem__(self, key, value):
+        return setattr(self, key, value)
+
+    def __contains__(self, key):
+        return key in self.__dict__
+
+    def __repr__(self):
+        return self.__dict__.__repr__()

webui.py

@@ -0,0 +1,70 @@
+import argparse
+import gradio as gr
+from gradio import components
+import os
+import torch
+import commons
+import utils
+from models import SynthesizerTrn
+from text.symbols import symbols
+from text import text_to_sequence
+from scipy.io.wavfile import write
+
+def get_text(text, hps):
+    text_norm = text_to_sequence(text, hps.data.text_cleaners)
+    if hps.data.add_blank:
+        text_norm = commons.intersperse(text_norm, 0)
+    text_norm = torch.LongTensor(text_norm)
+    return text_norm
+
+def tts(model_path, config_path, text):
+    model_path = './logs/' + model_path
+    config_path = './configs/' + config_path
+    hps = utils.get_hparams_from_file(config_path)
+
+    if "use_mel_posterior_encoder" in hps.model.keys() and hps.model.use_mel_posterior_encoder == True:
+        posterior_channels = 80
+        hps.data.use_mel_posterior_encoder = True
+    else:
+        posterior_channels = hps.data.filter_length // 2 + 1
+        hps.data.use_mel_posterior_encoder = False
+
+    net_g = SynthesizerTrn(
+        len(symbols),
+        posterior_channels,
+        hps.train.segment_size // hps.data.hop_length,
+        **hps.model).cuda()
+    _ = net_g.eval()
+    _ = utils.load_checkpoint(model_path, net_g, None)
+
+    stn_tst = get_text(text, hps)
+    x_tst = stn_tst.cuda().unsqueeze(0)
+    x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).cuda()
+
+    with torch.no_grad():
+        audio = net_g.infer(x_tst, x_tst_lengths, noise_scale=.667, noise_scale_w=0.8, length_scale=1)[0][0,0].data.cpu().float().numpy()
+
+    output_wav_path = "output.wav"
+    write(output_wav_path, hps.data.sampling_rate, audio)
+
+    return output_wav_path
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--model_path', type=str, default=None, help='Path to the model file.')
+    parser.add_argument('--config_path', type=str, default=None, help='Path to the config file.')
+    args = parser.parse_args()
+
+    model_files = [f for f in os.listdir('./logs/') if f.endswith('.pth')]
+    model_files.sort(key=lambda x: int(x.split('_')[-1].split('.')[0]), reverse=True)
+    config_files = [f for f in os.listdir('./configs/') if f.endswith('.json')]
+
+    default_model_file = args.model_path if args.model_path else (model_files[0] if model_files else None)
+    default_config_file = args.config_path if args.config_path else 'config.json'
+
+    gr.Interface(
+        fn=tts,
+        inputs=[components.Dropdown(model_files,value=default_model_file, label="Model File"), components.Dropdown(config_files,value=default_config_file, label="Config File"), components.Textbox(label="Text Input")],
+        outputs=components.Audio(type='filepath', label="Generated Speech"),
+        live=False
+    ).launch()