First model version

.flake8

@@ -0,0 +1,2 @@
+[flake8]
+per-file-ignores = __init__.py:F401

.gitignore

@@ -0,0 +1,21 @@
+**/.pre-commit-config.yaml
+**/*.ipynb
+*.wav
+*.npy
+*.pyc
+log
+*.png
+*.interval
+BZNSYP
+ckpt
+log
+nohup.out
+obsolete
+output
+output_no_hanzi
+results
+samples
+synth
+**/.ipynb_checkpoints/**
+**/__pycache__/
+mtts.egg-info/

.gitmodules

@@ -0,0 +1,12 @@
+[submodule "mtts/models/vocoder/VocGAN"]
+	path = mtts/models/vocoder/VocGAN
+	url = https://github.com/ranchlai/VocGAN.git
+[submodule "mtts/models/vocoder/hifi_gan"]
+	path = mtts/models/vocoder/hifi_gan
+	url = https://github.com/ranchlai/hifi-gan.git
+[submodule "mtts/models/vocoder/waveglow"]
+	path = mtts/models/vocoder/waveglow
+	url = https://github.com/ranchlai/waveglow.git
+[submodule "mtts/models/vocoder/melgan"]
+	path = mtts/models/vocoder/melgan
+	url = https://github.com/ranchlai/melgan.git

.style.yapf

@@ -0,0 +1,3 @@
+[style]
+based_on_style = pep8
+column_limit = 120

CONFIG.md

@@ -0,0 +1,54 @@
+# Introduction to configurations
+
+## Dataset
+``` yaml
+dataset: # the dataset part is for training only
+  train:
+    wav_scp: './train/wav.scp'
+    mel_scp: './train/mel.scp'
+    dur_scp: './train/dur.scp'
+    emb_type1:
+      _name: 'pinyin'
+      scp: './train/py.scp'
+      vocab: 'py.vocab'
+    emb_type2:
+      _name: 'graphic'
+      scp: './train/gp.scp'
+      vocab: 'gp.vocab'
+    #emb_type3:
+      #_name: 'speaker'
+     # scp: './train/spk.scp'
+     # vocab: # dosn't need vocab
+    emb_type4:
+      _name: 'prosody'
+      scp: './train/psd.scp'
+      vocab:
+```
+
+## Vocoder
+```yaml
+vocoder:
+  type: VocGan # choose one of the following
+  MelGAN:
+    checkpoint: ~/checkpoints/melgan/melgan_ljspeech.pth
+    config: ~/checkpoints/melgan/default.yaml
+    device: cpu
+  VocGan:
+    checkpoint: ~/checkpoints/vctk_pretrained_model_3180.pt #~/checkpoints/ljspeech_29de09d_4000.pt
+    denoise: True
+    device: cpu
+  HiFiGAN:
+    checkpoint: ~/checkpoints/VCTK_V3/generator_v3  # you need to download checkpoint and set the params here
+    device: cpu
+  Waveglow:
+    checkpoint:  ~/checkpoints/waveglow_256channels_universal_v5_state_dict.pt
+    sigma: 1.0
+    denoiser_strength: 0.0 # try 0.1
+    device: cpu #try cpu if out of memory
+
+```
+
+
+## Make your own changes
+Two config files are provided in the examples for illustration purpose. You can changed the config file if you know what you are doing.
+For example, you can remove speaker_emb from the following section, or add  prosody embedding if you have prosody label (as in biaobei dataset).

docs/_config.yml

@@ -0,0 +1 @@
+theme: jekyll-theme-hacker

docs/index.md

@@ -0,0 +1,547 @@
+
+
+
+## Audio samples
+
+Below are the audio samples from step 300k, pinyin + hanzi + unet(as postnet).
+
+It may take some time to load all audios in browser.
+
+
+### multi-speaker samples
+<table>
+   <tbody>
+      <tr>
+         <td><audio controls="controls">
+          <source src="./samples/aishell3/SSB0241_1.0_有个小孩叫小杜.上街打醋又买布.买了布打了醋.回头看见鹰抓兔.放.wav" autoplay="">
+        </audio>
+        </td>
+            <td><audio controls="controls">
+          <source src="./samples/aishell3/SSB0341_1.0_红鲤鱼绿鲤鱼与驴.wav" autoplay="">
+        </audio>
+        </td>
+      </tr>
+       <tr>
+         <td><audio controls="controls">
+          <source src="./samples/aishell3/SSB0395_1.0_这个小孩儿真懂事儿.wav" autoplay="">
+        </audio>
+        </td>
+         <td><audio controls="controls">
+          <source src="./samples/aishell3/SSB0535_1.0_前方路口左转，然后在下一个路口右转.wav" autoplay="">
+        </audio>
+        </td>
+      </tr>
+       <tr>
+         <td><audio controls="controls">
+          <source src="./samples/aishell3/SSB0606_1.0_天青色等烟雨,而我在等你,炊烟袅袅升起，隔江千万里.wav" autoplay="">
+        </audio>
+        </td>
+         <td><audio controls="controls">
+          <source src="./samples/aishell3/SSB0716_1.0_黑化肥发灰会挥发,灰化肥挥发会发黑.wav" autoplay="">
+        </audio>
+        </td>
+      </tr>
+       <tr>
+         <td><audio controls="controls">
+          <source src="./samples/aishell3/SSB1135_1.0_请输入您的卡号.您输入的卡号是六二二六,三八七六,零三四七,六九.wav" autoplay="">
+        </audio>
+        </td>
+           <td><audio controls="controls">
+          <source src="./samples/aishell3/SSB1402_1.0_这次考试有点儿紧张，下次再加油吧。.wav" autoplay="">
+        </audio>
+        </td>
+      </tr>
+       <tr>
+      <td><audio controls="controls">
+          <source src="./samples/aishell3/SSB1431_1.0_小毛豆你好，很高兴认识你,寒假马上要结束了.我觉得你最近有进步，.wav" autoplay="">
+        </audio>
+        </td>
+      </tr>
+   </tbody>
+</table>
+
+
+### Samples from training loop
+
+
+<table>
+   <thead>
+      <tr>
+         <th style="text-align: left">Synthesized</th>
+         <th style="text-align: left">Groudtruth</th>
+      </tr>
+   </thead>
+   <tbody>
+      <tr>
+         <td><audio controls="controls">
+          <source src="./data/step_300000_0_postnet_waveglow.wav" autoplay="">
+        </audio>
+        </td>
+         <td><audio controls="controls">
+          <source src="./data/step_300000_0_ground-truth_waveglow.wav" autoplay="">
+        </audio></td>
+      </tr>
+      <tr>
+         <td><audio controls="controls">
+          <source src="./data/step_300000_1_postnet_waveglow.wav" autoplay="">
+        </audio>
+        </td>
+         <td><audio controls="controls">
+          <source src="./data/step_300000_1_ground-truth_waveglow.wav" autoplay="">
+        </audio></td>
+      </tr>
+     <tr>
+         <td><audio controls="controls">
+          <source src="./data/step_300000_2_postnet_waveglow.wav" autoplay="">
+        </audio>
+        </td>
+         <td><audio controls="controls">
+          <source src="./data/step_300000_2_ground-truth_waveglow.wav" autoplay="">
+        </audio></td>
+      </tr>
+      <tr>
+         <td><audio controls="controls">
+          <source src="./data/step_300000_3_postnet_waveglow.wav" autoplay="">
+        </audio>
+        </td>
+         <td><audio controls="controls">
+          <source src="./data/step_300000_3_ground-truth_waveglow.wav" autoplay="">
+        </audio></td>
+      </tr>
+      <tr>
+         <td><audio controls="controls">
+          <source src="./data/step_300000_4_postnet_waveglow.wav" autoplay="">
+        </audio>
+        </td>
+         <td><audio controls="controls">
+          <source src="./data/step_300000_4_ground-truth_waveglow.wav" autoplay="">
+        </audio></td>
+      </tr>
+      <tr>
+         <td><audio controls="controls">
+          <source src="./data/step_300000_5_postnet_waveglow.wav" autoplay="">
+        </audio>
+        </td>
+         <td><audio controls="controls">
+          <source src="./data/step_300000_5_ground-truth_waveglow.wav" autoplay="">
+        </audio></td>
+      </tr>
+   </tbody>
+</table>
+
+
+
+### Novel sampels
+
+The following samples are generated by running
+```
+./scripts/hz_synth.sh 1.0 500000
+
+```
+
+<b> NOTE: there are many failure cases that are not shown here. </b>
+
+#### 常规句子 (注意字与字之前的音隔）
+
+- 前方路口左转，然后在下一个路口右转
+- 请输入您的卡号.您输入的卡号是六二二六,三八七六,零三四七,六九幺五
+- 天青色等烟雨,而我在等你,炊烟袅袅升起，隔江千万里
+- 我听不清你在说什么，请大声一点
+- 如果觉得这个项目好，请手动加个星吧，感谢
+- 小毛豆你好，很高兴认识你,寒假马上要结束了.我觉得你最近有进步，继续加油.晚上要早点睡觉,因为明天要开学了.
+
+
+
+<table>
+   <thead>
+      <tr>
+         <th style="text-align: left">Text</th>
+         <th style="text-align: left">Normal</th>
+         <th style="text-align: left">Fast</th>
+         <th style="text-align: left">slow</th>
+      </tr>
+   </thead>
+   <tbody>
+           <tr>
+         <td>1
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel2/hz_1.0_500000_前方路口左转，然后在下一个路口右转.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_0.9_500000_前方路口左转，然后在下一个路口右转.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_1.1_500000_前方路口左转，然后在下一个路口右转.wav" autoplay="">
+        </audio></td>
+      </tr>
+         <tr>
+         <td>2
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel2/hz_1.0_500000_请输入您的卡号.您输入的卡号是六二二六,三八七六,零三四七,六九.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_0.9_500000_请输入您的卡号.您输入的卡号是六二二六,三八七六,零三四七,六九.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_1.1_500000_请输入您的卡号.您输入的卡号是六二二六,三八七六,零三四七,六九.wav" autoplay="">
+        </audio></td>
+      </tr>
+         <tr>
+         <td>3
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel2/hz_1.0_500000_天青色等烟雨,而我在等你,炊烟袅袅升起，隔江千万里.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_0.9_500000_天青色等烟雨,而我在等你,炊烟袅袅升起，隔江千万里.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_1.1_500000_天青色等烟雨,而我在等你,炊烟袅袅升起，隔江千万里.wav" autoplay="">
+        </audio></td>
+      </tr>
+         <tr>
+         <td>4
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel2/hz_1.0_500000_我听不清你在说什么，请大声一点.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_0.9_500000_我听不清你在说什么，请大声一点.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_1.1_500000_我听不清你在说什么，请大声一点.wav" autoplay="">
+        </audio></td>
+      </tr>
+         <tr>
+         <td>5
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel2/hz_1.0_500000_如果觉得这个项目好，请手动加个星吧，感谢.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_0.9_500000_如果觉得这个项目好，请手动加个星吧，感谢.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_1.1_500000_如果觉得这个项目好，请手动加个星吧，感谢.wav" autoplay="">
+        </audio></td>
+      </tr>
+       <tr>
+         <td>6
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel2/hz_1.0_500000_小毛豆你好，很高兴认识你.寒假马上要结束了.我觉得你最近有进步，.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_0.9_500000_小毛豆你好，很高兴认识你,寒假马上要结束了.我觉得你最近有进步，.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_1.1_500000_小毛豆你好，很高兴认识你.寒假马上要结束了.我觉得你最近有进步，.wav" autoplay="">
+        </audio></td>
+      </tr>
+
+
+
+   </tbody>
+</table>
+#### 饶口令
+
+- 有个小孩叫小杜,上街打醋又买布.买了布打了醋,回头看见鹰抓兔.放下布搁下醋,上前去追鹰和兔.飞了鹰跑了兔,洒了醋湿了布.
+- 红鲤鱼绿鲤鱼与驴
+- 黑化肥发灰会挥发,灰化肥挥发会发黑
+
+
+
+<table>
+   <thead>
+      <tr>
+         <th style="text-align: left">Text</th>
+         <th style="text-align: left">Normal</th>
+         <th style="text-align: left">Fast</th>
+         <th style="text-align: left">slow</th>
+      </tr>
+   </thead>
+   <tbody>
+      <tr>
+         <td>1
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel2/hz_1.0_500000_黑化肥发灰会挥发,灰化肥挥发会发黑.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_0.9_500000_黑化肥发灰会挥发,灰化肥挥发会发黑.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_1.1_500000_黑化肥发灰会挥发,灰化肥挥发会发黑.wav" autoplay="">
+        </audio></td>
+      </tr>
+          <tr>
+         <td>2
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel2/hz_1.0_500000_红鲤鱼绿鲤鱼与驴.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_0.9_500000_红鲤鱼绿鲤鱼与驴.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_1.1_500000_红鲤鱼绿鲤鱼与驴.wav" autoplay="">
+        </audio></td>
+      </tr>
+       <tr>
+      <td>9
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel2/hz_1.0_500000_有个小孩叫小杜,上街打醋又买布.买了布打了醋,回头看见鹰抓兔.放.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_0.9_500000_有个小孩叫小杜,上街打醋又买布.买了布打了醋,回头看见鹰抓兔.放.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_1.1_500000_有个小孩叫小杜,上街打醋又买布.买了布打了醋,回头看见鹰抓兔.放.wav" autoplay="">
+        </audio></td>
+      </tr>
+
+
+   </tbody>
+</table>
+
+
+#### 儿化音
+
+在汉字转拼音时自动识别儿化音
+
+- 自个儿有多少斤两,心里还没数吗
+- 在这儿做啥呢,不能啥玩意儿都带到这里来?
+- 这个小孩儿真懂事儿
+<table>
+   <thead>
+      <tr>
+         <th style="text-align: left">Text</th>
+         <th style="text-align: left">Normal</th>
+         <th style="text-align: left">Fast</th>
+         <th style="text-align: left">slow</th>
+      </tr>
+   </thead>
+   <tbody>
+      <tr>
+         <td>1
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel2/hz_1.0_500000_自个儿有多少斤两,心里还没数吗.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_0.9_500000_自个儿有多少斤两,心里还没数吗.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_1.1_500000_自个儿有多少斤两,心里还没数吗.wav" autoplay="">
+        </audio></td>
+      </tr>
+          <tr>
+         <td>2
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel2/hz_1.0_500000_在这儿做啥呢,不能啥玩意儿都带到这里来?.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_0.9_500000_在这儿做啥呢,不能啥玩意儿都带到这里来?.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_1.1_500000_在这儿做啥呢,不能啥玩意儿都带到这里来?.wav" autoplay="">
+        </audio></td>
+      </tr>
+       <tr>
+      <td>9
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel2/hz_1.0_500000_这个小孩儿真懂事儿.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_0.9_500000_这个小孩儿真懂事儿.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel2/hz_1.1_500000_这个小孩儿真懂事儿.wav" autoplay="">
+        </audio></td>
+      </tr>
+
+
+   </tbody>
+</table>
+
+
+
+## The following is obsolete. 以下例子是旧版本，仅供参考
+
+### Novel synthsized samples
+
+#### Input text
+1. 黑化肥发灰会挥发,灰化肥挥发会发黑
+2. 红鲤鱼绿鲤鱼与驴
+3. 前方路口左转，然后在下一个路口右转
+4. 请输入您的卡号.您输入的卡号是六二二六,三八七六,零三四七,六九一五
+5. 天青色等烟雨,而我在等你,炊烟袅袅升起，隔江千万里
+6. 我听不清你在说什么，请大声一点
+7. 如果觉得这个项目好，请手动加个星吧，感谢
+8. 小毛豆你好，很高兴认识你.寒假马上要结束了.我觉得你最近有进步，继续加油.晚上要早点睡觉,因为明天要开学了.
+9. 有个小孩叫小杜,上街打醋又买布.买了布打了醋,回头看见鹰抓兔.放下布搁下醋,上前去追鹰和兔.飞了鹰跑了兔,洒了醋湿了布.
+
+
+#### Audio samples
+
+To generate audio samples, first you need to down load the checkpoint from <a href="https://drive.google.com/file/d/11mBus5gn69_KwvNec9Zy9jjTs3LgHdx3/view?usp=sharing">google drive</a> an untar it to mandarin_tts/
+
+
+Then the audio samples are genreated by the following command:
+```
+./scripts/hz_synth.sh 1.0 300000
+./scripts/hz_synth.sh 0.9 300000
+./scripts/hz_synth.sh 1.1 300000
+```
+
+You can also use pure pinyin + unet model, as follows:
+```
+./scripts/py_synth.sh 1.0 300000
+./scripts/py_synth.sh 0.9 300000
+./scripts/py_synth.sh 1.1 300000
+```
+
+
+For other text inputs, you can alter text in input.txt. Some hints:
+- 只需输入中文，因为代码中使用了pypinyin进行转换，转换前先分词。
+- 不同句子用句号分开，可放在同一行，最后同一行的句子会合并成一个语音。如果需要不同语音，放在不同行即可。
+- 需要停顿的地方用逗号。
+- 算法不区分问号和叹号，当成句号处理。同样，不区分顿号，当成逗号，token统一设置为sp1.
+
+
+<table>
+   <thead>
+      <tr>
+         <th style="text-align: left">Text</th>
+         <th style="text-align: left">Normal</th>
+         <th style="text-align: left">Fast</th>
+         <th style="text-align: left">slow</th>
+      </tr>
+   </thead>
+   <tbody>
+      <tr>
+         <td>1
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel/hz_1.0_300000_黑化肥发灰会挥发,灰化肥挥发会发黑.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_0.9_300000_黑化肥发灰会挥发,灰化肥挥发会发黑.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_1.1_300000_黑化肥发灰会挥发,灰化肥挥发会发黑.wav" autoplay="">
+        </audio></td>
+      </tr>
+          <tr>
+         <td>2
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel/hz_1.0_300000_红鲤鱼绿鲤鱼与驴.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_0.9_300000_红鲤鱼绿鲤鱼与驴.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_1.1_300000_红鲤鱼绿鲤鱼与驴.wav" autoplay="">
+        </audio></td>
+      </tr>
+        <tr>
+         <td>3
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel/hz_1.0_300000_前方路口左转，然后在下一个路口右转.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_0.9_300000_前方路口左转，然后在下一个路口右转.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_1.1_300000_前方路口左转，然后在下一个路口右转.wav" autoplay="">
+        </audio></td>
+      </tr>
+         <tr>
+         <td>4
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel/hz_1.0_300000_请输入您的卡号.您输入的卡号是六二二六,三八七六,零三四七,六九.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_0.9_300000_请输入您的卡号.您输入的卡号是六二二六,三八七六,零三四七,六九.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_1.1_300000_请输入您的卡号.您输入的卡号是六二二六,三八七六,零三四七,六九.wav" autoplay="">
+        </audio></td>
+      </tr>
+         <tr>
+         <td>5
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel/hz_1.0_300000_天青色等烟雨,而我在等你,炊烟袅袅升起，隔江千万里.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_0.9_300000_天青色等烟雨,而我在等你,炊烟袅袅升起，隔江千万里.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_1.1_300000_天青色等烟雨,而我在等你,炊烟袅袅升起，隔江千万里.wav" autoplay="">
+        </audio></td>
+      </tr>
+         <tr>
+         <td>6
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel/hz_1.0_300000_我听不清你在说什么，请大声一点.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_0.9_300000_我听不清你在说什么，请大声一点.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_1.1_300000_我听不清你在说什么，请大声一点.wav" autoplay="">
+        </audio></td>
+      </tr>
+         <tr>
+         <td>7
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel/hz_1.0_300000_如果觉得这个项目好，请手动加个星吧，感谢.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_0.9_300000_如果觉得这个项目好，请手动加个星吧，感谢.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_1.1_300000_如果觉得这个项目好，请手动加个星吧，感谢.wav" autoplay="">
+        </audio></td>
+      </tr>
+       <tr>
+         <td>8
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel/hz_1.0_300000_小毛豆你好，很高兴认识你.寒假马上要结束了.我觉得你最近有进步，.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_0.9_300000_小毛豆你好，很高兴认识你.寒假马上要结束了.我觉得你最近有进步，.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_1.1_300000_小毛豆你好，很高兴认识你.寒假马上要结束了.我觉得你最近有进步，.wav" autoplay="">
+        </audio></td>
+      </tr>
+    <tr>
+      <td>9
+        </td>
+         <td><audio controls="controls">
+          <source src="./novel/hz_1.0_300000_有个小孩叫小杜,上街打醋又买布.买了布打了醋,回头看见鹰抓兔.放.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_0.9_300000_有个小孩叫小杜,上街打醋又买布.买了布打了醋,回头看见鹰抓兔.放.wav" autoplay="">
+        </audio></td>
+          <td><audio controls="controls">
+          <source src="./novel/hz_1.1_300000_有个小孩叫小杜,上街打醋又买布.买了布打了醋,回头看见鹰抓兔.放.wav" autoplay="">
+        </audio></td>
+      </tr>
+
+
+   </tbody>
+</table>

examples/aishell3/config.yaml

@@ -0,0 +1,167 @@
+
+dataset:
+  train:
+    wav_scp: './train/wav.scp'
+    mel_scp: './train/mel.scp'
+    dur_scp: './train/dur.scp'
+    emb_type1:
+      _name: 'pinyin'
+      scp: './train/py.scp'
+      vocab: 'py.vocab'
+    emb_type2:
+      _name: 'graphic'
+      scp: './train/gp.scp'
+      vocab: 'gp.vocab'
+    emb_type3:
+      _name: 'speaker'
+      scp: './train/spk.scp'
+      vocab: ~ # dosn't need vocab
+    # NOTE: you can add more embedding here without changing the code.
+  eval:
+    # NOTE: this is not used for now, i.e., just training, no evaluation.
+    # You can use synthesize.py to check the training goes well, for now.
+
+
+
+training:
+  batch_size: 1024
+  batch_split: 64
+
+  epochs: 100000
+  grad_clip_thresh: 1.0
+  acc_step: 1
+  checkpoint_path: "./checkpoints/"
+  log_path: "./log/"
+  checkpoint_step: 5000
+  synth_step: 5000
+  log_step: 20
+  num_workers: 8
+
+  evaluation_step: 1000
+
+optimizer: # NOTE: if use SGD, params should change too, as it has different arguments.
+    type: Adam
+    n_warm_up_step: 2000
+    lr_decrease_step: 10000
+    lr_decrease_factor:
+    params:
+      betas: [0.9,0.98]
+      eps: !!float 1e-9
+      weight_decay: !!float 0.0
+      lr: !!float  1e-4
+lr_scheduler:
+    type: CyclicLR
+    params:
+      base_lr: !!float 1e-8
+      max_lr: !!float 1e-6
+      step_size_up: 5000
+      step_size_down: 5000
+      cycle_momentum: False
+
+fbank: # this is used for wav2mel.py
+  sample_rate: 22050
+  n_fft: 1024
+  hop_length: 256
+  win_length: 1024
+  max_wav_value: 32768.0
+  n_mels: 80
+  fmin: 0.0
+  fmax: 8000.0 # should be 11025 ?
+  mel_mean: -6.0304103
+
+encoder:
+    encoder_type: 'FS2TransformerEncoder'
+    conf:
+      n_layers: 4
+      n_heads: 2
+      hidden_dim: 256
+      dropout: 0.25
+      d_inner: 1024
+      max_len: 2048
+
+decoder:
+    decoder_type: 'FS2TransformerDecoder'
+    input_dim: 256 # should be the same as the output of encoder
+    n_layers: 4
+    n_heads: 2
+    hidden_dim: 256
+    d_inner: 1024
+    dropout: 0.25
+    max_len: 2048 # max len of seq, for position embedding pre-computation
+
+#(class) Decoder(input_dim: int = 256, n_layers: int = 4, n_heads: int = 2,
+#hidden_dim: int = 256, d_inner: int = 1024, dropout: float = 0.5, max_len: int = 2048
+postnet:
+  postnet_type: 'PostUNet' # 'PostUNet', 'PostNet1d'
+
+speaker_embedding:
+    enable: True
+    vocab: #None
+    vocab_size: 218 # aishell3 has 218 speakers
+    weight: 1.0 # you can play with weight here
+    dim: 256
+
+utterence_embedding:
+    enable: False # not implemented
+    type: 'lstm' # resnet
+    feature_config:
+      type: 'mel'
+      n_mels: 80
+      sampling_rate: 22050
+      n_fft: 1024
+      hop_length: 256
+      win_length: 1024
+
+hanzi_embedding:
+    enable: True
+    type: embedding
+    vocab: './gp.vocab'
+    dim: 256
+    weight: 0.5 # you can play with weight here
+    max_seq_len: 100
+
+pinyin_embedding:
+    enable: True
+    type: embedding
+    vocab: './py.vocab'
+    dim: 256
+    weight: 1.0
+    max_seq_len: 100
+
+duration_predictor:
+  input_dim: 256 # should be the same as encoder hiddien_dim
+  filter_size: 256
+  kernel_size: 3
+  dropout: 0.15 # important to set dropout here
+  duration_mean: 21.517294924096635 #for aishell3
+
+f0_predictor:
+  enable: False # currently not supported
+  filter_size: 256
+  kernel_size: 3
+  dropout: 0.5
+  n_bins: 256
+
+vocoder:
+  type: VocGan # choose one of the following
+  MelGAN:
+    checkpoint: ~/checkpoints/melgan/melgan_ljspeech.pth
+    config: ~/checkpoints/melgan/default.yaml
+    device: cpu
+  VocGan:
+    checkpoint: ~/checkpoints/vctk_pretrained_model_3180.pt #~/checkpoints/ljspeech_29de09d_4000.pt
+    denoise: True
+    device: cpu
+  HiFiGAN:
+    checkpoint: ~/checkpoints/VCTK_V3/generator_v3  # you need to download checkpoint and set the params here
+    device: cpu
+  Waveglow:
+    checkpoint:  ~/checkpoints/waveglow_256channels_universal_v5_state_dict.pt
+    sigma: 1.0
+    denoiser_strength: 0.0 # try 0.1
+    device: cpu #try cpu if out of memory
+
+
+
+synthesis:
+  normalize: True # normalize the sound volume

examples/aishell3/gp.vocab

@@ -0,0 +1,4502 @@
+sil
+一
+丁
+七
+万
+丈
+三
+三儿
+上
+下
+不
+与
+丐
+丑
+专
+且
+丕
+世
+丘
+丙
+业
+丛
+东
+丝
+丞
+丢
+两
+严
+丧
+个
+个儿
+丫
+中
+丰
+串
+串儿
+临
+丸
+丹
+为
+主
+丽
+丽儿
+举
+乃
+久
+么
+义
+之
+乌
+乍
+乎
+乏
+乐
+乐儿
+乒
+乓
+乔
+乖
+乘
+乙
+九
+乞
+也
+习
+乡
+书
+买
+乱
+乳
+乾
+了
+予
+争
+事
+事儿
+二
+于
+亏
+云
+互
+五
+井
+亚
+些
+亟
+亡
+亢
+交
+亦
+产
+亨
+亩
+享
+京
+亭
+亮
+亲
+亳
+亵
+人
+人儿
+亿
+什
+仁
+仁儿
+仄
+仅
+仆
+仇
+今
+今儿
+介
+仍
+从
+仑
+仓
+仔
+仕
+他
+仗
+付
+仙
+仞
+代
+令
+以
+仨
+仪
+们
+们儿
+仰
+仲
+件
+件儿
+价
+价儿
+任
+份
+份儿
+仿
+企
+伊
+伍
+伎
+伏
+伐
+休
+众
+优
+伙
+伙儿
+会
+会儿
+伞
+伟
+传
+伤
+伦
+伪
+伯
+估
+伴
+伴儿
+伶
+伸
+伺
+似
+伽
+佃
+但
+位
+位儿
+低
+住
+佐
+佑
+体
+何
+佘
+余
+佛
+作
+佟
+你
+你儿
+佣
+佩
+佬
+佯
+佰
+佳
+佼
+使
+侃
+侄
+來
+侈
+例
+侍
+侗
+供
+依
+侠
+侣
+侥
+侦
+侧
+侨
+侬
+侮
+侯
+侵
+便
+促
+俄
+俊
+俏
+俐
+俑
+俗
+俘
+保
+俞
+俟
+信
+俨
+俩
+俪
+俭
+修
+俱
+俵
+俺
+倍
+倍儿
+們
+倒
+倔
+倘
+候
+候儿
+倚
+借
+倡
+倦
+倩
+倪
+债
+值
+倾
+假
+偈
+偎
+偏
+做
+停
+健
+偶
+偷
+偿
+傅
+傍
+储
+催
+傲
+傻
+像
+僚
+僧
+僵
+僻
+儒
+儿
+兀
+允
+元
+兄
+充
+兆
+先
+光
+克
+克儿
+免
+兑
+兔
+党
+兜
+入
+全
+八
+公
+六
+兮
+兰
+共
+关
+兴
+兵
+其
+具
+典
+兹
+养
+兼
+兽
+冀
+内
+冈
+冉
+册
+再
+冒
+冕
+冗
+写
+军
+农
+冠
+冢
+冤
+冥
+冬
+冯
+冰
+冲
+决
+况
+冶
+冷
+冻
+冽
+净
+凄
+准
+准儿
+凇
+凉
+凋
+凌
+减
+凑
+凛
+凝
+几
+凡
+凤
+凭
+凯
+凰
+凳
+凶
+凸
+凹
+出
+击
+函
+凿
+刀
+刁
+刃
+分
+分儿
+切
+刈
+刊
+刑
+划
+列
+刘
+则
+刚
+创
+初
+删
+判
+刨
+利
+别
+刮
+到
+制
+刷
+刷儿
+券
+券儿
+刹
+刺
+刻
+刽
+剁
+剂
+剃
+削
+前
+剐
+剑
+剔
+剖
+剥
+剧
+剩
+剪
+副
+割
+剿
+劈
+力
+劝
+办
+功
+加
+务
+劣
+动
+助
+努
+劫
+劭
+励
+劲
+劲儿
+劳
+劵
+劵儿
+势
+勃
+勇
+勉
+勋
+勒
+勘
+募
+勤
+勺
+勺儿
+勾
+勿
+匀
+包
+匆
+匈
+匕
+化
+北
+匙
+匠
+匡
+匣
+匪
+匮
+匹
+区
+医
+匾
+匿
+十
+千
+升
+午
+午儿
+卉
+半
+半儿
+华
+华儿
+协
+卑
+卓
+卓儿
+单
+卖
+南
+博
+博儿
+卜
+占
+卡
+卢
+卤
+卦
+卧
+卫
+卯
+印
+危
+即
+却
+卵
+卷
+卷儿
+卸
+卿
+厂
+厄
+厅
+历
+厉
+压
+厌
+厕
+厘
+厚
+原
+厢
+厥
+厦
+厨
+厩
+厮
+去
+县
+参
+又
+叉
+及
+友
+友儿
+双
+反
+发
+叔
+取
+受
+变
+叙
+叛
+叠
+口
+口儿
+古
+句
+另
+叨
+叩
+只
+叫
+召
+叭
+叮
+可
+可儿
+台
+叱
+史
+右
+叶
+号
+号儿
+司
+叹
+叼
+吁
+吃
+各
+合
+吉
+吊
+同
+名
+名儿
+后
+吏
+吐
+向
+吒
+吓
+吕
+吗
+君
+吝
+吞
+吟
+否
+吧
+吧儿
+吨
+吩
+含
+听
+启
+吱
+吴
+吵
+吸
+吹
+吻
+吼
+吾
+呀
+呃
+呆
+呈
+告
+告儿
+呐
+呕
+呗
+员
+员儿
+呛
+呜
+呢
+呦
+周
+呱
+味
+味儿
+呵
+呼
+命
+咀
+咆
+咋
+和
+咎
+咏
+咐
+咒
+咖
+咘
+咙
+咚
+咤
+咧
+咨
+咪
+咫
+咬
+咭
+咯
+咱
+咳
+咸
+咽
+哀
+品
+哄
+哇
+哈
+哉
+响
+哎
+哑
+哒
+哔
+哗
+哟
+哥
+哦
+哨
+哩
+哪
+哪儿
+哭
+哮
+哲
+哺
+哼
+哽
+唇
+唉
+唐
+唔
+唠
+唤
+唬
+售
+唯
+唰
+唱
+唾
+啄
+商
+啊
+啜
+啡
+啤
+啥
+啦
+啧
+啪
+啰
+啸
+啼
+啾
+喀
+喂
+善
+喆
+喇
+喉
+喊
+喋
+喔
+喘
+喜
+喝
+喧
+喱
+喵
+喷
+喻
+喽
+嗄
+嗅
+嗑
+嗒
+嗓
+嗜
+嗝
+嗝儿
+嗡
+嗦
+嗨
+嗬
+嗯
+嗽
+嘀
+嘈
+嘉
+嘘
+嘛
+嘞
+嘟
+嘱
+嘲
+嘴
+嘿
+噜
+噢
+器
+噩
+噪
+噬
+噱
+噻
+嚎
+嚏
+嚣
+嚷
+嚼
+囊
+囍
+囔
+囚
+四
+回
+因
+团
+囤
+囧
+园
+园儿
+困
+围
+囷
+固
+国
+图
+图儿
+圃
+圆
+圈
+圈儿
+土
+圣
+在
+圩
+圭
+地
+圳
+场
+圾
+址
+坂
+均
+坊
+坍
+坎
+坎儿
+坏
+坐
+坑
+块
+块儿
+坚
+坛
+坝
+坞
+坟
+坠
+坡
+坤
+坦
+坨
+坪
+坯
+坷
+坻
+垂
+垃
+垄
+型
+型儿
+垒
+垚
+垛
+垟
+垠
+垡
+垣
+垤
+垦
+垫
+垮
+埂
+埃
+埋
+城
+埔
+埗
+埝
+域
+埠
+埭
+培
+基
+堂
+堃
+堆
+堆儿
+堕
+堡
+堤
+堪
+堰
+堵
+堺
+塌
+塑
+塔
+塘
+塞
+填
+塾
+境
+墅
+墉
+墓
+墙
+增
+墟
+墨
+墩
+墩儿
+壁
+壕
+壤
+士
+士儿
+壮
+声
+壳
+壳儿
+壶
+壹
+处
+备
+复
+夏
+夕
+夕儿
+外
+夙
+多
+夜
+够
+大
+天
+天儿
+太
+夫
+央
+夯
+失
+头
+头儿
+夷
+夸
+夹
+夺
+奂
+奇
+奈
+奉
+奋
+奎
+奏
+契
+奔
+奕
+奖
+套
+奠
+奢
+奥
+女
+女儿
+奴
+奶
+奸
+她
+好
+好儿
+如
+妃
+妄
+妆
+妇
+妇儿
+妈
+妊
+妍
+妒
+妓
+妖
+妙
+妞
+妞儿
+妥
+妨
+妪
+妮
+妳
+妹
+妻
+妾
+姆
+姊
+始
+姍
+姐
+姑
+姓
+委
+姗
+姚
+姜
+姣
+姥
+姨
+姬
+姻
+姿
+威
+娃
+娄
+娅
+娆
+娇
+娘
+娘儿
+娜
+娟
+娠
+娥
+娱
+娴
+娶
+娼
+婆
+婆儿
+婉
+婊
+婕
+婚
+婧
+婪
+婴
+婴儿
+婶
+婷
+婺
+婿
+媒
+媚
+媛
+媲
+媳
+媺
+嫁
+嫂
+嫉
+嫌
+嫖
+嫚
+嫡
+嫣
+嫦
+嫩
+嬅
+嬉
+孀
+子
+孔
+孕
+字
+存
+孙
+孚
+孜
+孝
+孟
+孢
+季
+孤
+学
+孩
+孩儿
+孪
+孬
+孵
+孺
+孽
+宁
+它
+宅
+宇
+守
+安
+宋
+完
+宏
+宕
+宗
+官
+宙
+宙儿
+定
+宛
+宜
+宝
+实
+宠
+审
+客
+客儿
+宣
+室
+宥
+宪
+宫
+宰
+害
+宴
+宵
+家
+家儿
+宸
+容
+宽
+宾
+宿
+寂
+寄
+寅
+密
+寇
+富
+寐
+寒
+寓
+寝
+寞
+察
+寡
+寤
+寥
+寨
+寮
+寸
+对
+寺
+寻
+导
+寿
+封
+射
+将
+尉
+尊
+小
+小儿
+少
+尔
+尖
+尘
+尚
+尝
+尤
+尧
+尬
+就
+尴
+尸
+尹
+尺
+尼
+尽
+尾
+尿
+局
+屁
+层
+居
+屈
+屉
+届
+屋
+屎
+屏
+屑
+展
+属
+屠
+屡
+履
+屯
+屯儿
+山
+屹
+屿
+岁
+岂
+岌
+岐
+岔
+岖
+岗
+岚
+岛
+岩
+岫
+岭
+岱
+岳
+岷
+岸
+峙
+峡
+峥
+峦
+峨
+峪
+峭
+峯
+峰
+峻
+崂
+崇
+崎
+崔
+崖
+崛
+崩
+崭
+崴
+嵇
+嵊
+嵋
+嵌
+嵩
+巅
+巍
+川
+川儿
+州
+巡
+巢
+工
+左
+巧
+巨
+巩
+巫
+差
+己
+已
+巴
+巷
+巾
+巿
+币
+市
+市儿
+布
+帅
+帆
+师
+希
+帐
+帕
+帖
+帘
+帜
+帝
+带
+带儿
+席
+帮
+帷
+常
+帼
+帽
+幂
+幄
+幅
+幌
+幕
+幢
+干
+干儿
+平
+年
+并
+幸
+幺
+幻
+幼
+幼儿
+幽
+广
+庄
+庆
+庆儿
+庇
+床
+序
+庐
+库
+应
+底
+底儿
+店
+店儿
+庙
+庙儿
+庚
+府
+庞
+废
+度
+座
+庭
+庵
+庶
+康
+庸
+庾
+廉
+廊
+廓
+廖
+廪
+延
+廷
+建
+开
+异
+弃
+弄
+弈
+弊
+弋
+式
+弑
+弓
+引
+弗
+弘
+弛
+弟
+张
+弥
+弦
+弧
+弯
+弱
+弹
+强
+弼
+归
+当
+录
+彝
+形
+彤
+彦
+彩
+彪
+彬
+彭
+彰
+影
+影儿
+彷
+役
+彻
+彼
+往
+征
+径
+待
+很
+徊
+律
+後
+徐
+徒
+得
+得儿
+徘
+御
+徨
+循
+微
+德
+徽
+心
+必
+忆
+忌
+忍
+忏
+志
+忘
+忙
+忠
+忡
+忤
+忧
+快
+忱
+念
+忻
+忽
+怀
+态
+怂
+怎
+怏
+怒
+怕
+怖
+怜
+思
+怠
+怡
+急
+性
+怨
+怪
+怯
+总
+怿
+恃
+恋
+恍
+恐
+恒
+恕
+恙
+恢
+恣
+恤
+恨
+恩
+恪
+恭
+息
+恰
+恳
+恶
+恺
+恻
+恼
+恿
+悄
+悉
+悍
+悔
+悔儿
+悖
+悚
+悟
+悠
+患
+悦
+您
+悬
+悱
+悲
+悴
+悼
+情
+惊
+惋
+惑
+惕
+惜
+惟
+惠
+惦
+惧
+惨
+惩
+惫
+惬
+惭
+惮
+惯
+惰
+想
+惶
+惹
+惺
+愁
+愈
+愉
+意
+意儿
+愕
+愚
+愛
+感
+愣
+愤
+愧
+愿
+慈
+慌
+慎
+慑
+慕
+慢
+慧
+慨
+慰
+慷
+憋
+憎
+憔
+憨
+憷
+憾
+懂
+懈
+懊
+懋
+懒
+懦
+懵
+懿
+戈
+戈儿
+戍
+戎
+戏
+成
+我
+戒
+或
+战
+戚
+戛
+戟
+截
+戬
+戮
+戳
+戴
+户
+户儿
+戾
+房
+所
+所儿
+扁
+扇
+扉
+手
+手儿
+才
+扎
+扑
+扒
+打
+扔
+托
+托儿
+扛
+扞
+扣
+扦
+执
+扩
+扫
+扬
+扭
+扮
+扯
+扰
+扳
+扶
+批
+扼
+找
+承
+技
+抄
+把
+抑
+抒
+抓
+投
+抖
+抗
+折
+抚
+抛
+抠
+抡
+抢
+护
+报
+抨
+披
+抬
+抱
+抵
+抹
+押
+抽
+拂
+拄
+担
+拆
+拇
+拈
+拉
+拌
+拍
+拎
+拐
+拒
+拓
+拔
+拖
+拗
+拘
+拙
+招
+招儿
+拜
+拟
+拢
+拣
+拥
+拦
+拧
+拨
+择
+括
+拭
+拮
+拯
+拱
+拳
+拴
+拷
+拼
+拽
+拾
+拿
+持
+挂
+指
+按
+挎
+挑
+挖
+挚
+挛
+挝
+挞
+挟
+挠
+挡
+挣
+挤
+挥
+挨
+挪
+挫
+振
+挺
+挽
+捂
+捅
+捆
+捉
+捍
+捎
+捏
+捐
+捕
+捘
+捞
+损
+捡
+换
+捣
+捧
+据
+捶
+捷
+掀
+授
+掉
+掌
+掏
+掐
+排
+掖
+掘
+掠
+探
+掣
+接
+控
+推
+掩
+措
+掰
+掳
+掷
+掺
+揍
+描
+提
+插
+握
+揣
+揪
+揭
+援
+揽
+搀
+搁
+搅
+搏
+搐
+搔
+搜
+搞
+搡
+搪
+搬
+搭
+携
+摁
+摄
+摆
+摇
+摊
+摊儿
+摔
+摘
+摞
+摧
+摩
+摸
+摹
+撂
+撅
+撇
+撑
+撒
+撕
+撞
+撤
+撩
+撬
+播
+撮
+撰
+撵
+撸
+撼
+擂
+擅
+操
+擎
+擒
+擘
+擞
+擦
+攀
+攒
+攘
+支
+收
+攸
+改
+攻
+放
+政
+故
+效
+敌
+敏
+救
+敖
+教
+敛
+敞
+敢
+散
+敦
+敬
+数
+数儿
+敲
+整
+敷
+文
+文儿
+斋
+斌
+斐
+斑
+斗
+料
+料儿
+斛
+斜
+斟
+斤
+斥
+斧
+斩
+断
+斯
+新
+方
+於
+施
+旁
+旅
+旋
+族
+旖
+旗
+无
+既
+日
+旦
+旧
+旨
+早
+早儿
+旬
+旭
+旮
+旯
+旱
+时
+旷
+旺
+昀
+昂
+昆
+昊
+昌
+明
+明儿
+昏
+易
+昔
+昕
+星
+映
+春
+春儿
+昧
+昨
+昭
+是
+是儿
+昱
+昵
+昼
+显
+時
+晃
+晋
+晏
+晒
+晓
+晔
+晕
+晖
+晗
+晚
+晞
+晞儿
+晟
+晤
+晦
+晨
+普
+景
+晰
+晴
+晶
+晷
+智
+晾
+暂
+暇
+暑
+暖
+暗
+暧
+暨
+暮
+暴
+曈
+曙
+曝
+曦
+曲
+曳
+更
+曹
+曼
+曾
+替
+最
+會
+月
+月儿
+有
+朋
+服
+朔
+朕
+朗
+望
+朝
+期
+朦
+木
+未
+末
+本
+本儿
+札
+术
+朱
+朴
+朵
+机
+朽
+杀
+杂
+权
+杆
+杆儿
+杉
+李
+杏
+杏儿
+材
+村
+村儿
+杖
+杜
+杞
+束
+杠
+条
+条儿
+来
+杨
+杭
+杯
+杰
+杷
+松
+板
+板儿
+极
+构
+枇
+枉
+枋
+析
+枕
+林
+枚
+果
+果儿
+枝
+枢
+枣
+枪
+枫
+枭
+枯
+枱
+架
+枷
+枸
+柄
+柏
+某
+柑
+染
+柔
+柚
+柜
+柠
+查
+查儿
+柬
+柯
+柱
+柳
+柴
+柿
+栀
+栅
+标
+标儿
+栈
+栋
+栏
+树
+栓
+栖
+栗
+校
+栩
+株
+栲
+样
+样儿
+核
+根
+根儿
+格
+格儿
+栽
+桂
+桃
+框
+案
+桉
+桌
+桎
+桐
+桑
+桓
+桔
+桠
+档
+桥
+桦
+桩
+桶
+梁
+梅
+梆
+梓
+梗
+梢
+梦
+梧
+梨
+梭
+梯
+械
+梳
+梵
+检
+棉
+棋
+棍
+棍儿
+棒
+棕
+棘
+棚
+棠
+森
+棱
+棵
+棺
+椅
+植
+椎
+椒
+椰
+楚
+楠
+楷
+楸
+楼
+概
+榄
+榆
+榈
+榔
+榕
+榜
+榨
+榫
+榭
+榴
+榷
+槌
+槐
+槛
+槛儿
+槟
+槭
+槽
+樊
+樟
+模
+模儿
+横
+樱
+橄
+橇
+橘
+橙
+橡
+橱
+檀
+檬
+欠
+次
+欢
+欣
+欧
+欲
+欸
+欺
+款
+歇
+歉
+歌
+歌儿
+歙
+止
+正
+此
+步
+武
+歧
+歪
+歹
+死
+殃
+殉
+殊
+残
+殖
+殡
+殴
+段
+段儿
+殷
+殿
+毁
+毅
+毋
+母
+每
+毒
+毓
+比
+毕
+毗
+毙
+毛
+毛儿
+毫
+毯
+氏
+民
+气
+氛
+氢
+氧
+氨
+氮
+氯
+氰
+水
+水儿
+永
+汀
+汁
+汁儿
+求
+汇
+汉
+汐
+汕
+汗
+汛
+汝
+江
+池
+污
+汤
+汨
+汪
+汰
+汶
+汽
+汾
+沁
+沂
+沃
+沅
+沈
+沉
+沌
+沐
+沒
+沓
+沙
+沛
+沟
+没
+沣
+沥
+沦
+沧
+沨
+沪
+沫
+沭
+沮
+沱
+河
+沸
+油
+油儿
+治
+沼
+沽
+沾
+沿
+泄
+泉
+泊
+泌
+泓
+泔
+法
+法儿
+泗
+泛
+泡
+波
+波儿
+泣
+泥
+注
+泪
+泫
+泯
+泰
+泳
+泳儿
+泵
+泷
+泸
+泻
+泼
+泽
+泾
+洁
+洋
+洒
+洗
+洙
+洛
+洞
+津
+洪
+洮
+洱
+洲
+洵
+活
+活儿
+洼
+洽
+派
+派儿
+流
+流儿
+浅
+浆
+浇
+测
+济
+浏
+浑
+浒
+浓
+浔
+浙
+浚
+浠
+浣
+浦
+浩
+浪
+浪儿
+浮
+浴
+海
+浸
+涂
+涅
+消
+涉
+涌
+涕
+涛
+涟
+涠
+涡
+涤
+润
+涧
+涨
+涩
+涪
+涮
+涯
+液
+涵
+涿
+淀
+淄
+淆
+淇
+淋
+淌
+淑
+淘
+淝
+淡
+淤
+淫
+淬
+淮
+深
+淳
+混
+淹
+添
+淼
+渀
+清
+渊
+渌
+渍
+渎
+渐
+渑
+渔
+渗
+渝
+渠
+渡
+渣
+渤
+渥
+温
+渭
+港
+渲
+渴
+游
+渺
+湃
+湉
+湍
+湖
+湘
+湛
+湾
+湿
+溃
+溅
+溆
+溉
+源
+源儿
+溜
+溜儿
+溢
+溦
+溧
+溪
+溯
+溶
+溺
+滁
+滇
+滋
+滑
+滔
+滕
+滚
+滞
+满
+滢
+滤
+滥
+滦
+滨
+滩
+滴
+漂
+漆
+漉
+漏
+漓
+演
+漕
+漠
+漩
+漪
+漫
+漭
+漯
+漱
+漳
+漾
+潇
+潋
+潍
+潘
+潜
+潢
+潭
+潮
+潼
+澄
+澍
+澎
+澜
+澡
+澳
+激
+濂
+濑
+濒
+濠
+濮
+濯
+瀑
+瀚
+瀛
+灌
+灏
+灞
+火
+灭
+灯
+灰
+灵
+灶
+灸
+灼
+灾
+灿
+炅
+炉
+炎
+炒
+炖
+炙
+炜
+炫
+炬
+炭
+炮
+炮儿
+炯
+炳
+炸
+点
+点儿
+炼
+炽
+烀
+烁
+烂
+烂儿
+烈
+烊
+烘
+烙
+烛
+烟
+烤
+烦
+烧
+烨
+烩
+烫
+热
+烯
+烷
+烹
+烽
+焉
+焊
+焓
+焕
+焖
+焘
+焚
+焦
+焯
+焰
+焱
+然
+煌
+煎
+煜
+煞
+煤
+煦
+照
+照儿
+煨
+煮
+煲
+煽
+熄
+熊
+熏
+熔
+熙
+熟
+熨
+熬
+熹
+燃
+燊
+燎
+燕
+燥
+爆
+爪
+爬
+爱
+爵
+父
+爷
+爸
+爸儿
+爹
+爽
+片
+片儿
+版
+版儿
+牌
+牌儿
+牙
+牛
+牛儿
+牟
+牡
+牢
+牧
+物
+牲
+牵
+特
+特儿
+牺
+犀
+犁
+犄
+犒
+犬
+犯
+状
+犸
+犹
+狂
+狄
+狐
+狗
+狗儿
+狙
+狞
+狠
+狡
+狩
+独
+狭
+狮
+狰
+狱
+狸
+狼
+猎
+猖
+猛
+猜
+猝
+猥
+猩
+猪
+猫
+献
+猴
+猷
+猾
+獒
+獗
+玄
+率
+玉
+王
+王儿
+玖
+玛
+玟
+玥
+玩
+玩儿
+玫
+玮
+环
+环儿
+现
+玲
+玷
+玹
+玺
+玻
+珀
+珂
+珉
+珊
+珍
+珏
+珑
+珞
+珠
+珠儿
+班
+班儿
+球
+球儿
+琅
+理
+琉
+琊
+琐
+琛
+琢
+琥
+琦
+琨
+琪
+琳
+琴
+琼
+瑄
+瑕
+瑙
+瑛
+瑜
+瑞
+瑟
+瑰
+瑶
+瑾
+璀
+璃
+璇
+璐
+璞
+璧
+璨
+瓜
+瓣
+瓦
+瓮
+瓯
+瓶
+瓶儿
+瓷
+甄
+甘
+甚
+甜
+生
+生儿
+甥
+用
+甩
+甫
+甭
+甯
+田
+由
+甲
+申
+电
+男
+甸
+町
+画
+画儿
+畅
+界
+畏
+畔
+留
+畜
+略
+番
+畳
+畴
+畸
+疃
+疆
+疏
+疑
+疖
+疗
+疗儿
+疚
+疝
+疤
+疫
+疯
+疱
+疲
+疵
+疸
+疹
+疼
+疾
+病
+症
+痉
+痊
+痒
+痕
+痘
+痛
+痞
+痪
+痫
+痰
+痱
+痴
+痹
+瘀
+瘟
+瘠
+瘢
+瘤
+瘦
+瘪
+瘫
+瘳
+瘾
+癌
+癖
+癫
+登
+白
+百
+皂
+的
+的儿
+皆
+皇
+皋
+皑
+皓
+皖
+皙
+皮
+皱
+皿
+盂
+盆
+盈
+益
+盏
+盐
+监
+盒
+盒儿
+盔
+盖
+盗
+盘
+盛
+盟
+目
+目儿
+盯
+盱
+盲
+直
+相
+盹儿
+盼
+盾
+省
+眉
+看
+眙
+真
+眠
+眨
+眩
+眯
+眶
+眷
+眸
+眼
+眼儿
+着
+睁
+睐
+睛
+睡
+睢
+督
+睦
+睫
+睬
+睹
+睽
+睿
+瞄
+瞅
+瞌
+瞎
+瞑
+瞒
+瞥
+瞧
+瞩
+瞪
+瞬
+瞳
+瞻
+瞿
+矗
+矛
+矜
+矢
+矣
+知
+矩
+矫
+短
+矮
+石
+矶
+矿
+码
+砂
+砌
+砍
+研
+砖
+砚
+砣
+砭
+砰
+破
+砸
+砺
+础
+硅
+硕
+硚
+硝
+硫
+硬
+确
+碌
+碍
+碎
+碑
+碗
+碘
+碚
+碜
+碟
+碧
+碰
+碱
+碳
+碴
+碾
+磁
+磅
+磊
+磋
+磐
+磕
+磨
+磷
+礁
+示
+礼
+社
+社儿
+祁
+祈
+祉
+祐
+祖
+祛
+祝
+祝儿
+神
+祟
+祠
+祢
+祥
+票
+祭
+祯
+祷
+祸
+祺
+禀
+禁
+禄
+禅
+福
+禧
+禹
+禺
+离
+禽
+禾
+秀
+私
+秃
+秆
+秆儿
+秉
+秋
+种
+科
+科儿
+秒
+秘
+租
+秤
+秦
+秩
+积
+称
+秸
+移
+秽
+稀
+程
+稍
+税
+稚
+稣
+稳
+稻
+稼
+稽
+稿
+稿儿
+穆
+穗
+穴
+究
+穷
+空
+穿
+突
+窃
+窄
+窈
+窍
+窑
+窒
+窕
+窖
+窗
+窘
+窜
+窝
+窟
+窥
+窦
+立
+竖
+站
+竞
+竟
+章
+竣
+童
+竭
+端
+竹
+竺
+竿
+竿儿
+笃
+笆
+笈
+笋
+笑
+笔
+笙
+笛
+笠
+符
+笨
+第
+笳
+笺
+笼
+等
+筋
+筐
+筑
+筒
+答
+策
+筛
+筝
+筠
+筱
+筵
+筷
+筹
+签
+签儿
+简
+箍
+算
+管
+管儿
+箫
+箭
+箱
+篆
+篇
+篑
+篓
+篡
+篮
+篱
+篷
+篼
+簇
+簧
+簿
+籁
+籍
+米
+类
+籽
+粉
+粑
+粒
+粒儿
+粕
+粗
+粘
+粟
+粤
+粥
+粪
+粮
+粱
+粹
+精
+糊
+糕
+糖
+糗
+糙
+糜
+糟
+糠
+糯
+系
+紊
+素
+索
+紧
+紫
+累
+給
+絮
+繁
+纠
+纡
+红
+纣
+纤
+约
+级
+纪
+纫
+纬
+纭
+纯
+纱
+纲
+纳
+纵
+纶
+纷
+纸
+纹
+纺
+纽
+线
+练
+组
+绅
+细
+织
+终
+绊
+绍
+绎
+经
+绑
+绒
+结
+绕
+绘
+给
+绚
+绛
+络
+绝
+绞
+统
+绢
+绢儿
+绣
+绥
+继
+绩
+绪
+续
+绮
+绯
+绰
+绳
+维
+绵
+绷
+绸
+综
+绽
+绿
+缄
+缅
+缆
+缇
+缈
+缉
+缎
+缓
+缔
+缕
+编
+缘
+缚
+缜
+缝
+缠
+缢
+缤
+缥
+缨
+缩
+缪
+缬
+缭
+缮
+缰
+缴
+缸
+缺
+罂
+罄
+罐
+罐儿
+网
+罕
+罗
+罚
+罡
+罢
+罩
+罩儿
+罪
+置
+署
+羁
+羊
+美
+羔
+羚
+羞
+羡
+群
+羲
+羹
+羽
+翁
+翅
+翊
+翎
+翔
+翘
+翟
+翠
+翡
+翩
+翰
+翱
+翻
+翼
+耀
+老
+考
+者
+者儿
+而
+耍
+耐
+耒
+耕
+耗
+耘
+耙
+耳
+耶
+耷
+耸
+耻
+耽
+耿
+聂
+聆
+聊
+聋
+职
+联
+联儿
+聖
+聘
+聚
+聪
+聿
+肃
+肆
+肇
+肉
+肋
+肌
+肖
+肘
+肚
+肛
+肝
+肠
+股
+肢
+肤
+肥
+肩
+肪
+肮
+肯
+育
+肴
+肺
+肾
+肿
+胀
+胁
+胃
+胆
+胆儿
+背
+胎
+胖
+胗
+胚
+胛
+胜
+胞
+胡
+胤
+胥
+胧
+胭
+胯
+胰
+胳
+胶
+胸
+胺
+能
+脂
+脆
+脉
+脊
+脏
+脐
+脑
+脑儿
+脓
+脖
+脚
+脱
+脸
+脸儿
+脹
+脾
+腆
+腊
+腋
+腌
+腐
+腑
+腔
+腕
+腕儿
+腥
+腩
+腰
+腱
+腹
+腺
+腻
+腼
+腾
+腿
+腿儿
+膀
+膀儿
+膊
+膏
+膏儿
+膛
+膜
+膝
+膨
+膳
+膺
+臀
+臂
+臃
+臆
+臣
+臧
+自
+臭
+至
+致
+臻
+舀
+舅
+舆
+舌
+舍
+舒
+舔
+舜
+舞
+舟
+航
+舫
+般
+舰
+舱
+舵
+舶
+舷
+舸
+船
+艇
+艘
+良
+艰
+色
+艳
+艺
+艾
+节
+芊
+芋
+芍
+芒
+芓
+芙
+芜
+芝
+芥
+芦
+芩
+芪
+芬
+芭
+芮
+芯
+花
+花儿
+芳
+芸
+芹
+芽
+苇
+苋
+苍
+苏
+苑
+苓
+苗
+苛
+苟
+苡
+苣
+若
+苦
+苯
+英
+苷
+苹
+茁
+茂
+范
+范儿
+茄
+茅
+茉
+茎
+茗
+茜
+茨
+茫
+茬
+茬儿
+茱
+茵
+茶
+茸
+茹
+荃
+荆
+草
+草儿
+荐
+荒
+荔
+荚
+荟
+荠
+荡
+荣
+荤
+荧
+荨
+荫
+药
+荷
+荼
+莅
+莆
+莉
+莎
+莒
+莓
+莘
+莞
+莠
+莫
+莱
+莲
+莴
+获
+莹
+莺
+莽
+菁
+菇
+菊
+菌
+菏
+菜
+菠
+菡
+菩
+菱
+菲
+萃
+萄
+萌
+萍
+萎
+萝
+萤
+营
+萦
+萧
+萨
+萱
+萼
+落
+葆
+著
+葛
+葛儿
+葡
+董
+葩
+葫
+葬
+葱
+葵
+蒂
+蒋
+蒙
+蒜
+蒲
+蒸
+蒿
+蓄
+蓉
+蓓
+蓝
+蓬
+蓼
+蔑
+蔓
+蔚
+蔡
+蔬
+蔷
+蔽
+蕃
+蕉
+蕊
+蕙
+蕨
+蕴
+蕾
+薄
+薇
+薛
+薪
+薯
+薰
+薹
+藁
+藉
+藏
+藤
+藩
+藻
+蘑
+蘸
+虎
+虐
+虑
+虚
+虞
+虫
+虹
+虽
+虾
+蚀
+蚁
+蚂
+蚊
+蚌
+蚓
+蚕
+蚝
+蚤
+蚯
+蛇
+蛊
+蛋
+蛙
+蛛
+蛟
+蛮
+蛾
+蜀
+蜂
+蜃
+蜇
+蜍
+蜒
+蜓
+蜕
+蜗
+蜘
+蜚
+蜜
+蜡
+蜢
+蜷
+蜻
+蜿
+蝇
+蝉
+蝎
+蝙
+蝠
+蝴
+蝶
+螂
+螃
+螈
+融
+螨
+螫
+螯
+螳
+螵
+螺
+蟀
+蟋
+蟑
+蟒
+蟹
+蟾
+蠢
+血
+衅
+行
+行儿
+衍
+衔
+街
+衡
+衢
+衣
+补
+表
+表儿
+衫
+衫儿
+衬
+衮
+衰
+衷
+袁
+袂
+袄
+袈
+袋
+袋儿
+袍
+袒
+袖
+袜
+袤
+被
+袭
+袱
+裁
+裂
+装
+裆
+裔
+裕
+裘
+裙
+裙儿
+裟
+裤
+裱
+裴
+裸
+裹
+褂儿
+褐
+褓
+褔
+褛
+褥
+褪
+褴
+襁
+襄
+襟
+西
+要
+覆
+見
+见
+观
+规
+视
+览
+觉
+觎
+角
+角儿
+觞
+解
+触
+言
+詹
+誉
+誓
+警
+譬
+计
+订
+认
+讥
+讧
+讨
+让
+讪
+训
+议
+讯
+记
+讲
+讳
+讶
+许
+讹
+论
+讼
+讽
+设
+设儿
+访
+诀
+证
+证儿
+评
+诅
+识
+诈
+诉
+诊
+词
+译
+诓
+试
+诗
+诙
+诚
+诛
+话
+话儿
+诞
+诟
+诠
+诡
+询
+该
+详
+诧
+诩
+诫
+诬
+语
+误
+诱
+说
+说儿
+诵
+诶
+请
+诸
+诺
+读
+诽
+课
+诿
+谁
+调
+谅
+谆
+谈
+谊
+谋
+谌
+谍
+谎
+谏
+谐
+谑
+谓
+谕
+谙
+谚
+谜
+谢
+谣
+谤
+谦
+谨
+谩
+谬
+谭
+谱
+谱儿
+谴
+谷
+豁
+豆
+豌
+豚
+象
+豪
+豫
+豹
+貌
+贝
+贝儿
+贞
+负
+贡
+财
+责
+贤
+败
+账
+货
+质
+贩
+贪
+贫
+贬
+购
+贯
+贱
+贴
+贵
+贷
+贸
+费
+贺
+贻
+贼
+贾
+贿
+赁
+赂
+赃
+资
+赉
+赊
+赋
+赌
+赎
+赏
+赐
+赔
+赖
+赘
+赚
+赛
+赝
+赞
+赠
+赡
+赢
+赣
+赤
+赦
+赫
+走
+赴
+赵
+赶
+起
+趁
+超
+越
+趋
+趟
+趣
+足
+趴
+趸
+趾
+跃
+跆
+跋
+跌
+跑
+跑儿
+距
+跟
+跟儿
+跤
+跨
+跪
+路
+跳
+践
+跷
+跺
+跻
+踊
+踌
+踏
+踝
+踞
+踢
+踩
+踪
+踵
+踹
+蹄
+蹈
+蹋
+蹦
+蹭
+蹲
+蹴
+蹶
+蹿
+躁
+躇
+身
+躬
+躯
+躲
+躺
+车
+车儿
+轧
+轨
+轩
+转
+轮
+轮儿
+软
+轰
+轴
+轶
+轻
+轼
+载
+轿
+较
+辄
+辅
+辆
+辈
+辉
+辊
+辍
+辐
+辑
+输
+辖
+辗
+辙
+辛
+辜
+辞
+辟
+辣
+辨
+辩
+辫
+辰
+辱
+边
+边儿
+辽
+达
+达儿
+迁
+迄
+迅
+过
+迈
+迎
+运
+近
+返
+还
+这
+这儿
+进
+远
+违
+连
+迟
+迥
+迪
+迫
+迭
+述
+迷
+迸
+迹
+追
+退
+送
+适
+逃
+逅
+逆
+选
+逊
+逍
+透
+逐
+递
+途
+逗
+通
+逛
+逝
+逞
+速
+造
+逢
+逮
+逸
+逻
+逼
+逾
+遂
+遇
+遍
+遏
+道
+遗
+遛
+遢
+遣
+遥
+遭
+遮
+遴
+遵
+避
+邀
+邂
+邃
+邋
+邑
+邓
+邢
+那
+那儿
+邦
+邪
+邮
+邯
+邰
+邱
+邳
+邵
+邸
+邹
+邺
+邻
+郁
+郊
+郎
+郑
+郜
+郝
+郡
+郧
+部
+郭
+郴
+郸
+都
+鄂
+鄙
+鄚
+鄞
+酉
+酊
+酋
+酌
+配
+酒
+酗
+酝
+酣
+酥
+酩
+酪
+酬
+酮
+酯
+酱
+酵
+酶
+酷
+酸
+酿
+醇
+醉
+醋
+醒
+醛
+醺
+采
+采儿
+释
+里
+重
+野
+量
+金
+釜
+鉴
+銮
+鑫
+针
+钉
+钊
+钌
+钏
+钓
+钗
+钙
+钛
+钝
+钞
+钟
+钠
+钢
+钥
+钦
+钧
+钨
+钩
+钮
+钯
+钰
+钱
+钲
+钳
+钴
+钵
+钻
+钿
+铀
+铁
+铂
+铃
+铅
+铆
+铎
+铐
+铖
+铛
+铜
+铝
+铠
+铡
+铤
+铩
+铬
+铭
+铮
+铲
+银
+铸
+铺
+铺儿
+铼
+链
+链儿
+铿
+销
+锁
+锂
+锄
+锅
+锅儿
+锈
+锋
+锌
+锐
+错
+锚
+锜
+锡
+锣
+锤
+锥
+锦
+锭
+键
+锯
+锵
+锷
+锹
+锻
+镀
+镁
+镇
+镉
+镌
+镍
+镐
+镑
+镖
+镜
+镜儿
+镯
+镳
+镶
+长
+门
+门儿
+闪
+闫
+闭
+问
+闯
+闲
+间
+闵
+闷
+闸
+闹
+闺
+闻
+闽
+闾
+阀
+阁
+阂
+阅
+阆
+阎
+阐
+阑
+阔
+阕
+阙
+阜
+队
+队儿
+阡
+阪
+阮
+阱
+防
+阳
+阴
+阵
+阶
+阻
+阿
+陀
+陂
+附
+际
+陆
+陇
+陈
+陈儿
+陋
+陌
+降
+限
+陕
+陛
+陡
+院
+院儿
+除
+陨
+险
+陪
+陵
+陶
+陶儿
+陷
+隅
+隆
+隋
+隍
+随
+隐
+隔
+隙
+障
+隧
+隶
+隽
+难
+雀
+雁
+雄
+雅
+集
+雇
+雌
+雍
+雏
+雕
+雨
+雪
+雪儿
+雯
+雳
+零
+雷
+雾
+需
+霄
+霆
+震
+霉
+霍
+霏
+霓
+霖
+霜
+霞
+霪
+露
+霸
+霹
+霾
+靂
+青
+靓
+靖
+静
+靛
+非
+靠
+靡
+面
+面儿
+革
+靳
+靴
+靶
+鞋
+鞍
+鞘
+鞠
+鞭
+韦
+韧
+韩
+韬
+韭
+音
+韵
+韶
+页
+页儿
+顶
+顷
+项
+顺
+须
+顽
+顾
+顿
+颁
+颂
+预
+颅
+领
+颇
+颈
+颊
+颍
+颐
+频
+颓
+颖
+颗
+题
+颜
+额
+颠
+颢
+颤
+颦
+风
+飒
+飓
+飘
+飙
+飞
+食
+餐
+饥
+饨
+饪
+饭
+饮
+饰
+饱
+饲
+饵
+饶
+饺
+饼
+饿
+馄
+馅
+馅儿
+馆
+馆儿
+馈
+馋
+馍
+馏
+馒
+首
+首儿
+香
+馥
+馨
+马
+马儿
+驭
+驰
+驱
+驳
+驴
+驶
+驸
+驹
+驻
+驼
+驾
+驿
+骁
+骂
+骄
+骅
+骆
+骇
+验
+骏
+骑
+骗
+骚
+骜
+骝
+骞
+骡
+骤
+骤儿
+骥
+骨
+骰
+骷
+骸
+骼
+髅
+髋
+髓
+高
+髦
+鬃
+鬟
+鬼
+魁
+魂
+魄
+魅
+魏
+魔
+魟
+鮑
+鱼
+鱿
+鲁
+鲇
+鲍
+鲑
+鲜
+鲟
+鲢
+鲤
+鲨
+鲱
+鲶
+鲸
+鳄
+鳅
+鳌
+鳖
+鳝
+鳞
+鸟
+鸠
+鸡
+鸣
+鸥
+鸦
+鸪
+鸭
+鸯
+鸳
+鸵
+鸽
+鸿
+鹂
+鹅
+鹉
+鹊
+鹌
+鹏
+鹑
+鹜
+鹟
+鹤
+鹦
+鹧
+鹰
+鹿
+麋
+麒
+麓
+麟
+麦
+麻
+麽
+麾
+黄
+黎
+黏
+黑
+黔
+默
+黛
+黯
+鼎
+鼓
+鼠
+鼻
+齐
+齿
+龄
+龈
+龍
+龑
+龙
+龚
+龟

examples/aishell3/input.txt

@@ -0,0 +1,22 @@
+#name|py|gp|spk
+text1|sil ni3 qu4 zuo4 fan4 ba5 sil|sil 你 去 做 饭 吧 sil|0 
+text2|sil ni3 qu4 zuo4 fan4 ba5 sil|sil 你 去 做 饭 吧 sil|10 
+text3|sil ni3 qu4 zuo4 fan4 ba5 sil|sil 你 去 做 饭 吧 sil|0 10
+text4|sil ni3 qu4 zuo4 fan4 ba5 sil|sil 你 去 做 饭 吧 sil|100
+text5|sil ni3 qu4 zuo4 fan4 ba5 sil|sil 你 去 做 饭 吧 sil|33
+text6|sil ni3 qu4 zuo4 fan4 ba5 sil|sil 你 去 做 饭 吧 sil|66
+text3a|sil you3 ge4 xiao3 hai2 jiao4 xiao3 du4 sil|sil 有　个 小 孩 叫 小 杜 sil|100
+text4a|sil shang4 jie1 da3 cu4 you4 mai3 bu4 sil|sil 上 街 打 醋 又 买 布 sil|2
+text5a|sil mai3 le5 bu4 da3 le5 cu4 sil|sil 买 了 布 打 了 醋 sil|88
+text6a|sil hui2 tou2 kan4 jian4 ying1 zhua1 tu4 sil|sil 回 头 看 见 鹰 抓 兔 sil|40
+text7a|sil fang4 xia4 bu4 ge1 xia4 cu4 sil|sil 放 下 布 搁 下 醋 sil|25
+text8a|sil shang4 qian2 qu4 zhui1 ying1 he2 tu4 sil|sil 上 前 去 追 鹰 和 兔 sil|0
+text9a|sil fei1 le5 ying1 pao3 le5 tu4 sil|sil 飞 了 鹰 跑 了 兔 sil|106
+text10a|sil sa3 le5 cu4 shi1 le5 bu4 sil|sil 洒 了 醋 湿 了 布 sil|62
+
+#demonstrate how to mix two speaker to generate new sound!
+change1|sil wei4 shi4 ying4 xin1 de5 wang3 luo4 chuan2 bo1 fang1 shi4 he2 du2 zhe3 yue4 du2 xi2 guan4 sil|sil 为 适 应 新 的 网 络 传 播 方 式 和 读 者 阅 读 习 惯 sil|20 20 20 20
+change2|sil wei4 shi4 ying4 xin1 de5 wang3 luo4 chuan2 bo1 fang1 shi4 he2 du2 zhe3 yue4 du2 xi2 guan4 sil|sil 为 适 应 新 的 网 络 传 播 方 式 和 读 者 阅 读 习 惯 sil|20 20 20 88
+change3|sil wei4 shi4 ying4 xin1 de5 wang3 luo4 chuan2 bo1 fang1 shi4 he2 du2 zhe3 yue4 du2 xi2 guan4 sil|sil 为 适 应 新 的 网 络 传 播 方 式 和 读 者 阅 读 习 惯 sil|20 20 88 88
+change4|sil wei4 shi4 ying4 xin1 de5 wang3 luo4 chuan2 bo1 fang1 shi4 he2 du2 zhe3 yue4 du2 xi2 guan4 sil|sil 为 适 应 新 的 网 络 传 播 方 式 和 读 者 阅 读 习 惯 sil|20 88 88 88
+change5|sil wei4 shi4 ying4 xin1 de5 wang3 luo4 chuan2 bo1 fang1 shi4 he2 du2 zhe3 yue4 du2 xi2 guan4 sil|sil 为 适 应 新 的 网 络 传 播 方 式 和 读 者 阅 读 习 惯 sil|88 88 88 88  

examples/aishell3/prepare.py

@@ -0,0 +1,96 @@
+import glob
+import os
+import argparse
+
+from mtts.utils.logging import get_logger
+logger = get_logger(__file__)
+
+def augment_cn_with_sil(py_sent, cn_sent):
+    sil_loc = [i for i, p in enumerate(py_sent.split()) if p == 'sil']
+    han = [h for i, h in enumerate(cn_sent.split()) if h != 'sil']
+
+    k = 0
+    final = []
+    for i in range(len(han) + len(sil_loc)):
+        if i in sil_loc:
+            final += ['sil']
+        else:
+            final += [han[k]]
+            k += 1
+    return ' '.join(final)
+
+def write_scp(filename,scp):
+        with open(filename,'wt') as f:
+            f.write('\n'.join(scp)+'\n')
+
+if __name__ == '__main__':
+
+    parser = argparse.ArgumentParser(description='data pre-processing')
+    parser.add_argument('--meta_file', type=str, required=False, default='name_py_hz_dur.txt')
+    parser.add_argument('--wav_folder', type=str, required=False, default='./wavs')
+    parser.add_argument('--mel_folder', type=str, required=False, default='./mels')
+    parser.add_argument('--dst_folder', type=str, required=False, default='./train')
+    parser.add_argument('--generate_vocab', type=bool, required=False, default=False)
+    
+    args = parser.parse_args()
+
+    logger.info(args)
+
+    lines = open(args.meta_file).read().split('\n')
+    lines = [l.split('|') for l in lines if len(l)>0]
+    files = glob.glob(f'{args.wav_folder}/*.wav')
+    logger.info(f'{len(files)} wav files found')
+    lines0 = []
+    
+    for name,py,gp,dur in lines:
+        gp = augment_cn_with_sil(py,gp) # make sure gp and py has the same # of sil
+        assert len(py.split()) == len(gp.split()), f'error in {name}:{py},{gp}'
+        lines0 += [(name,py,gp,dur)]
+    lines = lines0
+
+    wav_scp = []
+    mel_scp = []
+    gp_scp = []
+    py_scp = []
+    dur_scp = []
+    spk_scp = []
+    all_spk = []
+    all_spk = [l[0][:7] for l in lines]
+    all_spk = list(set(all_spk))
+    all_spk.sort()
+    spk2idx = {s:i for i,s in enumerate(all_spk)}
+    all_py = []
+    all_gp = []
+
+    for name,py,gp,dur in lines:
+        
+        wav_scp += [name +' ' + f'{args.wav_folder}/{name}.wav']
+        mel_scp += [name +' ' +  f'{args.mel_folder}/{name}.npy']
+        py_scp += [name+' '+ py]
+        gp_scp += [name+' '+ gp]
+        dur_scp += [name+' '+ dur]
+        n = len(gp.split())
+        spk_idx = spk2idx[name[:7]]
+        spk_scp += [name + ' ' + ' '.join([str(spk_idx)]*n)]
+    if args.generate_vocab:
+        logger.warn('Caution: The vocab generated might be different from others(e.g., pretained models)')
+        pyvocab = list(set(all_py))
+        gpvocab = list(set(all_gp))
+        pyvocab.sort()
+        gpvocab.sort(key=lambda x: pypinyin.pinyin(x,0)[0][0])
+        with open('py.vocab','wt') as f:
+            f.write('\n'.join(pyvocab))
+        with open('gp.vocab','wt') as f:
+            f.write('\n'.join(gpvocab))
+
+
+    os.makedirs(args.dst_folder,exist_ok=True)
+    write_scp(f'{args.dst_folder}/wav.scp',wav_scp)
+    write_scp(f'{args.dst_folder}/py.scp',py_scp)
+    write_scp(f'{args.dst_folder}/gp.scp',gp_scp)
+    write_scp(f'{args.dst_folder}/dur.scp',dur_scp)
+    write_scp(f'{args.dst_folder}/spk.scp',spk_scp)
+    write_scp(f'{args.dst_folder}/mel.scp',mel_scp)
+
+
+

examples/aishell3/py.vocab

@@ -0,0 +1,1915 @@
+<unk>
+a1
+a2
+a3
+a4
+a5
+ai1
+ai2
+ai3
+ai4
+ai5
+an1
+an2
+an3
+an4
+ang1
+ang2
+ang4
+ao1
+ao2
+ao3
+ao4
+ba1
+ba2
+ba3
+ba4
+ba5
+bai1
+bai2
+bai3
+bai4
+bai5
+ban1
+ban2
+ban3
+ban4
+ban5
+bang1
+bang2
+bang3
+bang4
+bangr3
+banr1
+banr3
+banr4
+bao1
+bao2
+bao3
+bao4
+bao5
+bar4
+bar5
+bei1
+bei2
+bei3
+bei4
+bei5
+beir4
+ben1
+ben2
+ben3
+ben4
+ben5
+beng1
+beng2
+beng3
+beng4
+beng5
+benr3
+bi1
+bi2
+bi3
+bi4
+bi5
+bian1
+bian2
+bian3
+bian4
+bian5
+bianr1
+biao1
+biao2
+biao3
+biao4
+biaor1
+biaor3
+bie1
+bie2
+bie3
+bie4
+bie5
+bin1
+bin2
+bin3
+bin4
+bing1
+bing2
+bing3
+bing4
+bo1
+bo2
+bo3
+bo4
+bo5
+bor1
+bor2
+bu1
+bu2
+bu3
+bu4
+bu5
+ca1
+ca2
+ca4
+cai1
+cai2
+cai3
+cai4
+cai5
+cair3
+can1
+can2
+can3
+can4
+cang1
+cang2
+cang3
+cang4
+cao1
+cao2
+cao3
+cao4
+caor3
+ce1
+ce3
+ce4
+ce5
+cen1
+cen2
+cen4
+ceng1
+ceng2
+ceng3
+ceng4
+cha1
+cha2
+cha3
+cha4
+chai1
+chai2
+chai3
+chai4
+chan1
+chan2
+chan3
+chan4
+chan5
+chang1
+chang2
+chang3
+chang4
+chang5
+chao1
+chao2
+chao3
+chao4
+char2
+che1
+che2
+che3
+che4
+chen1
+chen2
+chen4
+chen5
+cheng1
+cheng2
+cheng3
+cheng4
+cheng5
+chenr2
+cher1
+chi1
+chi2
+chi3
+chi4
+chi5
+chong1
+chong2
+chong3
+chong4
+chong5
+chou1
+chou2
+chou3
+chou4
+chou5
+chu1
+chu2
+chu3
+chu4
+chu5
+chuai1
+chuai3
+chuai4
+chuan1
+chuan2
+chuan3
+chuan4
+chuang1
+chuang2
+chuang3
+chuang4
+chuanr1
+chuanr4
+chui1
+chui2
+chui4
+chui5
+chun1
+chun2
+chun3
+chun5
+chunr1
+chuo1
+chuo4
+ci1
+ci2
+ci3
+ci4
+ci5
+cong1
+cong2
+cong3
+cong4
+cong5
+cou1
+cou2
+cou4
+cu1
+cu2
+cu3
+cu4
+cu5
+cuan1
+cuan2
+cuan4
+cui1
+cui2
+cui3
+cui4
+cui5
+cun1
+cun2
+cun3
+cun4
+cun5
+cunr1
+cuo1
+cuo3
+cuo4
+da1
+da2
+da3
+da4
+da5
+dai1
+dai2
+dai3
+dai4
+dair4
+dan1
+dan2
+dan3
+dan4
+dan5
+dang1
+dang2
+dang3
+dang4
+danr3
+dao1
+dao2
+dao3
+dao4
+dao5
+dar2
+de1
+de2
+de3
+de4
+de5
+dei2
+dei3
+dei5
+den1
+den2
+den3
+deng1
+deng2
+deng3
+deng4
+deng5
+der2
+der5
+di1
+di2
+di3
+di4
+di5
+dian1
+dian2
+dian3
+dian4
+dian5
+dianr2
+dianr3
+dianr4
+dianr5
+diao1
+diao2
+diao4
+die1
+die2
+die3
+die4
+die5
+ding1
+ding2
+ding3
+ding4
+ding5
+dir3
+diu1
+diu4
+dong1
+dong2
+dong3
+dong4
+dong5
+dou1
+dou2
+dou3
+dou4
+dou5
+dour4
+du1
+du2
+du3
+du4
+du5
+duan1
+duan2
+duan3
+duan4
+duanr4
+dui1
+dui4
+duir1
+duir4
+dun1
+dun2
+dun3
+dun4
+dun5
+dunr1
+dunr3
+duo1
+duo2
+duo3
+duo4
+duo5
+e1
+e2
+e3
+e4
+e5
+ei1
+ei5
+en1
+en2
+en4
+en5
+eng2
+eng4
+er1
+er2
+er3
+er4
+er5
+fa1
+fa2
+fa3
+fa4
+fa5
+fan1
+fan2
+fan3
+fan4
+fan5
+fang1
+fang2
+fang3
+fang4
+fang5
+fangr4
+fanr4
+far2
+far3
+fei1
+fei2
+fei3
+fei4
+fen1
+fen2
+fen3
+fen4
+fen5
+feng1
+feng2
+feng3
+feng4
+feng5
+fenr1
+fenr4
+fenr5
+fo2
+fou2
+fou3
+fu1
+fu2
+fu3
+fu4
+fu5
+fur4
+fur5
+ga1
+ga4
+gai1
+gai2
+gai3
+gai4
+gan1
+gan2
+gan3
+gan4
+gang1
+gang2
+gang3
+gang4
+gang5
+ganr1
+ganr3
+gao1
+gao2
+gao3
+gao4
+gaor1
+gaor3
+gaor4
+ge1
+ge2
+ge3
+ge4
+ge5
+gei1
+gei2
+gei3
+gei5
+gen1
+gen3
+gen4
+geng1
+geng2
+geng3
+geng4
+genr1
+ger1
+ger2
+ger3
+ger4
+ger5
+gong1
+gong2
+gong3
+gong4
+gong5
+gou1
+gou2
+gou3
+gou4
+gou5
+gour3
+gu1
+gu2
+gu3
+gu4
+gu5
+gua1
+gua3
+gua4
+guai1
+guai2
+guai3
+guai4
+guan1
+guan2
+guan3
+guan4
+guang1
+guang2
+guang3
+guang4
+guanr2
+guanr3
+guanr4
+guar4
+gui1
+gui2
+gui3
+gui4
+gui5
+gun1
+gun2
+gun3
+gun4
+gunr4
+guo1
+guo2
+guo3
+guo4
+guo5
+guor1
+guor3
+ha1
+ha2
+ha4
+ha5
+hai1
+hai2
+hai3
+hai4
+hai5
+hair2
+han1
+han2
+han3
+han4
+han5
+hang1
+hang2
+hang3
+hang4
+hangr2
+hao1
+hao2
+hao3
+hao4
+hao5
+haor1
+haor3
+haor4
+haor5
+he1
+he2
+he3
+he4
+he5
+hei1
+hei2
+hei4
+hen1
+hen2
+hen3
+hen4
+heng1
+heng2
+heng3
+heng4
+heng5
+her2
+hong1
+hong2
+hong3
+hong4
+hou1
+hou2
+hou3
+hou4
+hou5
+hour4
+hour5
+hu1
+hu2
+hu3
+hu4
+hu5
+hua1
+hua2
+hua3
+hua4
+hua5
+huai1
+huai2
+huai3
+huai4
+huan1
+huan2
+huan3
+huan4
+huan5
+huang1
+huang2
+huang3
+huang4
+huanr2
+huar1
+huar2
+huar4
+hui1
+hui2
+hui3
+hui4
+hui5
+huir2
+huir3
+huir4
+hun1
+hun2
+hun3
+hun4
+hun5
+huo2
+huo3
+huo4
+huo5
+huor2
+huor3
+huor5
+hur4
+ji1
+ji2
+ji3
+ji4
+ji5
+jia1
+jia2
+jia3
+jia4
+jia5
+jian1
+jian2
+jian3
+jian4
+jian5
+jiang1
+jiang2
+jiang3
+jiang4
+jianr4
+jiao1
+jiao2
+jiao3
+jiao4
+jiar1
+jiar4
+jie1
+jie2
+jie3
+jie4
+jie5
+jin1
+jin2
+jin3
+jin4
+jin5
+jing1
+jing2
+jing3
+jing4
+jing5
+jingr4
+jinr1
+jinr4
+jiong1
+jiong2
+jiong3
+jiong4
+jiu1
+jiu2
+jiu3
+jiu4
+jiu5
+ju1
+ju2
+ju3
+ju4
+ju5
+juan1
+juan2
+juan3
+juan4
+juanr3
+juanr4
+jue1
+jue2
+jue3
+jue4
+jue5
+juer2
+juer3
+jun1
+jun3
+jun4
+jun5
+ka1
+ka2
+ka3
+ka4
+kai1
+kai2
+kai3
+kai4
+kai5
+kan1
+kan2
+kan3
+kan4
+kan5
+kang1
+kang2
+kang3
+kang4
+kanr3
+kao1
+kao2
+kao3
+kao4
+ke1
+ke2
+ke3
+ke4
+ke5
+ken1
+ken2
+ken3
+keng1
+keng2
+keng3
+ker1
+ker2
+ker3
+ker4
+kong1
+kong2
+kong3
+kong4
+kou1
+kou2
+kou3
+kou4
+kour3
+ku1
+ku2
+ku3
+ku4
+ku5
+kua1
+kua3
+kua4
+kuai3
+kuai4
+kuair4
+kuair5
+kuan1
+kuan2
+kuan3
+kuan4
+kuang1
+kuang2
+kuang3
+kuang4
+kuang5
+kui1
+kui2
+kui4
+kun1
+kun3
+kun4
+kuo4
+la1
+la2
+la3
+la4
+la5
+lai1
+lai2
+lai3
+lai4
+lai5
+lan1
+lan2
+lan3
+lan4
+lang2
+lang3
+lang4
+langr4
+lanr4
+lao1
+lao2
+lao3
+lao4
+lao5
+lar3
+le1
+le2
+le3
+le4
+le5
+lei1
+lei2
+lei3
+lei4
+lei5
+leng1
+leng2
+leng3
+leng4
+ler4
+li1
+li2
+li3
+li4
+li5
+lia2
+lia3
+lian1
+lian2
+lian3
+lian4
+lian5
+liang1
+liang2
+liang3
+liang4
+liang5
+lianr2
+lianr3
+lianr4
+liao1
+liao2
+liao3
+liao4
+liao5
+liaor2
+liaor4
+lie1
+lie2
+lie4
+lin1
+lin2
+lin3
+lin4
+ling1
+ling2
+ling3
+ling4
+ling5
+lir4
+liu1
+liu2
+liu3
+liu4
+liu5
+liur1
+liur2
+lo5
+long2
+long3
+long4
+lou1
+lou2
+lou3
+lou4
+lou5
+lu1
+lu2
+lu3
+lu4
+lu5
+luan2
+luan3
+luan4
+lun1
+lun2
+lun4
+lunr2
+luo1
+luo2
+luo3
+luo4
+luo5
+lv2
+lv3
+lv4
+lv5
+lve4
+ma1
+ma2
+ma3
+ma4
+ma5
+mai1
+mai2
+mai3
+mai4
+man1
+man2
+man3
+man4
+man5
+mang1
+mang2
+mang3
+mang4
+mao1
+mao2
+mao3
+mao4
+mao5
+maor2
+mar3
+me1
+me2
+me3
+me4
+me5
+mei1
+mei2
+mei3
+mei4
+mei5
+men1
+men2
+men3
+men4
+men5
+meng1
+meng2
+meng3
+meng4
+meng5
+menr2
+menr5
+mi1
+mi2
+mi3
+mi4
+mi5
+mian1
+mian2
+mian3
+mian4
+mian5
+mianr4
+miao1
+miao2
+miao3
+miao4
+miaor4
+mie2
+mie4
+min2
+min3
+min4
+min5
+ming1
+ming2
+ming3
+ming4
+ming5
+mingr2
+miu3
+miu4
+mo1
+mo2
+mo3
+mo4
+mo5
+mor2
+mou2
+mou3
+mu1
+mu2
+mu3
+mu4
+mu5
+mur4
+mur5
+na1
+na2
+na3
+na4
+na5
+nai2
+nai3
+nai4
+nai5
+nan1
+nan2
+nan3
+nan4
+nan5
+nang1
+nang2
+nang4
+nang5
+nao1
+nao2
+nao3
+nao4
+nao5
+naor2
+naor3
+nar2
+nar3
+nar4
+nar5
+ne1
+ne2
+ne4
+ne5
+nei1
+nei3
+nei4
+nen2
+nen4
+neng2
+neng3
+neng4
+neng5
+ni1
+ni2
+ni3
+ni4
+ni5
+nian1
+nian2
+nian3
+nian4
+nian5
+niang1
+niang2
+niang3
+niang4
+niang5
+niangr2
+niao1
+niao2
+niao3
+niao4
+nie1
+nie2
+nie4
+nin2
+nin3
+nin4
+ning1
+ning2
+ning3
+ning4
+nir3
+niu1
+niu2
+niu3
+niu4
+niu5
+niur1
+niur2
+nong2
+nong4
+nu1
+nu2
+nu3
+nu4
+nuan2
+nuan3
+nuan4
+nuo2
+nuo4
+nv2
+nv3
+nv4
+nv5
+nve4
+nvr3
+o1
+o2
+o5
+ou1
+ou2
+ou3
+ou4
+pa1
+pa2
+pa3
+pa4
+pai1
+pai2
+pai3
+pai4
+pai5
+pair2
+pair4
+pan1
+pan2
+pan4
+pang1
+pang2
+pang4
+pang5
+pao1
+pao2
+pao3
+pao4
+pao5
+paor3
+paor4
+pei1
+pei2
+pei3
+pei4
+pei5
+pen1
+pen2
+pen4
+peng1
+peng2
+peng3
+peng4
+peng5
+pi1
+pi2
+pi3
+pi4
+pian1
+pian2
+pian4
+pian5
+pianr1
+pianr4
+piao1
+piao2
+piao3
+piao4
+pie1
+pie2
+pie3
+pin1
+pin2
+pin3
+pin4
+pin5
+ping1
+ping2
+ping3
+ping4
+ping5
+pingr2
+po1
+po2
+po3
+po4
+po5
+por2
+pou1
+pu1
+pu2
+pu3
+pu4
+pu5
+pur1
+pur2
+qi1
+qi2
+qi3
+qi4
+qi5
+qia1
+qia2
+qia3
+qia4
+qian1
+qian2
+qian3
+qian4
+qian5
+qiang1
+qiang2
+qiang3
+qianr1
+qiao1
+qiao2
+qiao3
+qiao4
+qiao5
+qie1
+qie2
+qie3
+qie4
+qin1
+qin2
+qin3
+qin4
+qin5
+qing1
+qing2
+qing3
+qing4
+qing5
+qingr4
+qiong2
+qiu1
+qiu2
+qiu3
+qiu5
+qiur2
+qu1
+qu2
+qu3
+qu4
+qu5
+quan1
+quan2
+quan3
+quan4
+quan5
+quanr1
+quanr4
+que1
+que2
+que4
+qun2
+qun3
+qun4
+qunr2
+ran1
+ran2
+ran3
+ran4
+rang1
+rang2
+rang3
+rang4
+rang5
+rao2
+rao3
+rao4
+re2
+re3
+re4
+ren1
+ren2
+ren3
+ren4
+ren5
+reng1
+reng2
+reng3
+reng4
+renr2
+ri2
+ri4
+rong2
+rong3
+rong4
+rong5
+rou2
+rou3
+rou4
+ru1
+ru2
+ru3
+ru4
+ruan2
+ruan3
+ruan4
+rui2
+rui3
+rui4
+run2
+run3
+run4
+run5
+ruo1
+ruo4
+sa1
+sa2
+sa3
+sa4
+sa5
+sai1
+sai2
+sai3
+sai4
+san1
+san2
+san3
+san4
+sang1
+sang3
+sang4
+sanr1
+sao1
+sao2
+sao3
+sao4
+se2
+se4
+se5
+sen1
+sen2
+sen3
+sen4
+sen5
+seng1
+seng3
+seng4
+sha1
+sha2
+sha3
+sha4
+sha5
+shai1
+shai2
+shai3
+shai4
+shan1
+shan2
+shan3
+shan4
+shan5
+shang1
+shang2
+shang3
+shang4
+shang5
+shanr1
+shao1
+shao2
+shao3
+shao4
+shao5
+shaor2
+she1
+she2
+she3
+she4
+shei1
+shei2
+shei3
+shen1
+shen2
+shen3
+shen4
+shen5
+sheng1
+sheng2
+sheng3
+sheng4
+sheng5
+shengr1
+sher4
+shi1
+shi2
+shi3
+shi4
+shi5
+shir4
+shou1
+shou2
+shou3
+shou4
+shour3
+shu1
+shu2
+shu3
+shu4
+shu5
+shua1
+shua2
+shua3
+shua4
+shuai1
+shuai2
+shuai3
+shuai4
+shuan1
+shuan2
+shuan4
+shuang1
+shuang3
+shuang4
+shuar1
+shui1
+shui2
+shui3
+shui4
+shui5
+shuir2
+shuir3
+shun1
+shun3
+shun4
+shun5
+shuo1
+shuo2
+shuo3
+shuo4
+shuo5
+shuor1
+shur4
+si1
+si2
+si3
+si4
+si5
+sil
+song1
+song2
+song3
+song4
+sou1
+sou2
+sou3
+sou4
+sou5
+su1
+su2
+su3
+su4
+su5
+suan1
+suan2
+suan3
+suan4
+sui1
+sui2
+sui3
+sui4
+sun1
+sun2
+sun3
+sun4
+suo1
+suo2
+suo3
+suo4
+suo5
+suor3
+ta1
+ta2
+ta3
+ta4
+ta5
+tai1
+tai2
+tai3
+tai4
+tai5
+tan1
+tan2
+tan3
+tan4
+tan5
+tang1
+tang2
+tang3
+tang4
+tanr1
+tao1
+tao2
+tao3
+tao4
+tao5
+taor2
+te2
+te3
+te4
+te5
+teng1
+teng2
+teng4
+teng5
+ter4
+ti1
+ti2
+ti3
+ti4
+ti5
+tian1
+tian2
+tian3
+tian4
+tian5
+tianr1
+tiao1
+tiao2
+tiao3
+tiao4
+tiao5
+tiaor2
+tie1
+tie2
+tie3
+tie4
+ting1
+ting2
+ting3
+ting5
+tong1
+tong2
+tong3
+tong4
+tong5
+tou1
+tou2
+tou3
+tou4
+tou5
+tour2
+tu1
+tu2
+tu3
+tu4
+tu5
+tuan1
+tuan2
+tui1
+tui2
+tui3
+tui4
+tuir3
+tun1
+tun2
+tun4
+tunr2
+tuo1
+tuo2
+tuo3
+tuo4
+tuo5
+tuor1
+tur2
+wa1
+wa2
+wa3
+wa4
+wa5
+wai1
+wai2
+wai3
+wai4
+wan1
+wan2
+wan3
+wan4
+wang1
+wang2
+wang3
+wang4
+wang5
+wangr2
+wangr4
+wanr2
+wanr4
+wei1
+wei2
+wei3
+wei4
+wei5
+weir4
+wen1
+wen2
+wen3
+wen4
+wen5
+weng1
+weng4
+wenr2
+wo1
+wo2
+wo3
+wo4
+wo5
+wu1
+wu2
+wu3
+wu4
+wu5
+wur3
+xi1
+xi2
+xi3
+xi4
+xi5
+xia1
+xia2
+xia3
+xia4
+xia5
+xian1
+xian2
+xian3
+xian4
+xiang1
+xiang2
+xiang3
+xiang4
+xiang5
+xianr4
+xiao1
+xiao2
+xiao3
+xiao4
+xiao5
+xiaor3
+xie1
+xie2
+xie3
+xie4
+xie5
+xin1
+xin2
+xin4
+xin5
+xing1
+xing2
+xing3
+xing4
+xing5
+xingr2
+xingr4
+xiong1
+xiong2
+xiong4
+xir1
+xiu1
+xiu2
+xiu3
+xiu4
+xu1
+xu2
+xu3
+xu4
+xuan1
+xuan2
+xuan3
+xuan4
+xue1
+xue2
+xue3
+xue4
+xuer3
+xun1
+xun2
+xun3
+xun4
+ya1
+ya2
+ya3
+ya4
+ya5
+yan1
+yan2
+yan3
+yan4
+yang1
+yang2
+yang3
+yang4
+yang5
+yangr4
+yanr3
+yao1
+yao2
+yao3
+yao4
+yao5
+ye1
+ye2
+ye3
+ye4
+ye5
+yer4
+yi1
+yi2
+yi3
+yi4
+yi5
+yin1
+yin2
+yin3
+yin4
+yin5
+ying1
+ying2
+ying3
+ying4
+ying5
+yingr1
+yingr2
+yingr3
+yir4
+yo1
+yo3
+yong1
+yong2
+yong3
+yong4
+yong5
+yongr3
+you1
+you2
+you3
+you4
+you5
+your2
+your3
+your4
+yu1
+yu2
+yu3
+yu4
+yu5
+yuan1
+yuan2
+yuan3
+yuan4
+yuan5
+yuanr2
+yuanr4
+yue1
+yue2
+yue4
+yue5
+yuer4
+yun1
+yun2
+yun3
+yun4
+za1
+za2
+za3
+za4
+zai1
+zai2
+zai3
+zai4
+zai5
+zan1
+zan2
+zan3
+zan4
+zang1
+zang2
+zang3
+zang4
+zao1
+zao2
+zao3
+zao4
+zao5
+zaor3
+ze1
+ze2
+ze3
+ze4
+ze5
+zei2
+zei3
+zei4
+zen1
+zen2
+zen3
+zen4
+zeng1
+zeng2
+zeng3
+zeng4
+zha1
+zha2
+zha4
+zha5
+zhai1
+zhai2
+zhai3
+zhai4
+zhan1
+zhan2
+zhan3
+zhan4
+zhan5
+zhang1
+zhang2
+zhang3
+zhang4
+zhao1
+zhao2
+zhao3
+zhao4
+zhao5
+zhaor1
+zhaor4
+zhe1
+zhe2
+zhe3
+zhe4
+zhe5
+zhei4
+zhen1
+zhen2
+zhen3
+zhen4
+zhen5
+zheng1
+zheng2
+zheng3
+zheng4
+zheng5
+zhengr4
+zher3
+zher4
+zhi1
+zhi2
+zhi3
+zhi4
+zhi5
+zhir1
+zhong1
+zhong2
+zhong3
+zhong4
+zhou1
+zhou2
+zhou3
+zhou4
+zhour4
+zhu1
+zhu2
+zhu3
+zhu4
+zhu5
+zhua1
+zhua3
+zhuai1
+zhuai3
+zhuai4
+zhuan1
+zhuan2
+zhuan3
+zhuan4
+zhuang1
+zhuang3
+zhuang4
+zhuang5
+zhui1
+zhui3
+zhui4
+zhun1
+zhun2
+zhun3
+zhun4
+zhunr3
+zhuo1
+zhuo2
+zhuo3
+zhuo4
+zhuo5
+zhuor2
+zhur1
+zhur4
+zi1
+zi2
+zi3
+zi4
+zi5
+zong1
+zong2
+zong3
+zong4
+zou1
+zou2
+zou3
+zou4
+zu1
+zu2
+zu3
+zu4
+zuan1
+zuan3
+zuan4
+zui1
+zui2
+zui3
+zui4
+zun1
+zun3
+zun4
+zuo1
+zuo2
+zuo3
+zuo4
+zuo5

examples/biaobei/Wave/desktop.ini

@@ -0,0 +1,4 @@
+[ViewState]
+Mode=
+Vid=
+FolderType=Generic

examples/biaobei/config.yaml

@@ -0,0 +1,161 @@
+
+dataset:
+  train:
+    wav_scp: './train/wav.scp'
+    mel_scp: './train/mel.scp'
+    dur_scp: './train/dur.scp'
+    emb_type1:
+      _name: 'pinyin'
+      scp: './train/py.scp'
+      vocab: 'py.vocab'
+    emb_type2:
+      _name: 'graphic'
+      scp: './train/gp.scp'
+      vocab: 'gp.vocab'
+
+    emb_type3:
+      _name: 'speaker'
+      scp: './train/spk.scp'
+      vocab: # dosn't need vocab
+
+training:
+  batch_size: 16
+  batch_split: 1
+  epochs: 10000
+  grad_clip_thresh: 1.0
+  acc_step: 1
+  checkpoint_path: "./checkpoints/"
+  log_path: "./log/"
+  checkpoint_step: 5000
+  synth_step: 5000
+  log_step: 20
+  num_workers: 8
+
+  evaluation_step: 1000
+
+optimizer:
+    type: Adam
+    n_warm_up_step: 2000
+    #lr_decrease_step: 10000
+    #lr_decrease_factor:
+    params:
+      betas: [0.9,0.98]
+      eps: !!float 1e-9
+      weight_decay: !!float 0.0
+      lr: !!float  1e-4
+lr_scheduler:
+    type: CyclicLR
+    params:
+      base_lr: !!float 1e-7
+      max_lr: !!float 1e-4
+      step_size_up: 5000
+      step_size_down: 8000
+      cycle_momentum: False
+
+vocoder:
+  type: VocGan # choose one of the following
+  MelGAN:
+    checkpoint: ~/checkpoints/melgan/melgan_ljspeech.pth
+    config: ~/checkpoints/melgan/default.yaml
+    device: cpu
+  VocGan:
+    checkpoint: ~/checkpoints/vctk_pretrained_model_3180.pt #~/checkpoints/ljspeech_29de09d_4000.pt
+    denoise: True
+    device: cpu
+  HiFiGAN:
+    checkpoint: ~/checkpoints/VCTK_V3/generator_v3  # you need to download checkpoint and set the params here
+    device: cpu
+  Waveglow:
+    checkpoint:  ~/checkpoints/waveglow_256channels_universal_v5_state_dict.pt
+    sigma: 1.0
+    denoiser_strength: 0.0 # try 0.1
+    device: cpu #try cpu if out of memory
+
+fbank:
+  sample_rate: 16000
+  n_fft: 1024
+  hop_length: 256
+  win_length: 1024
+  max_wav_value: 32768.0
+  n_mels: 80
+  fmin: 0.0
+  fmax: 8000.0 # should be 11025
+  mel_mean: -6.0304103
+
+encoder:
+    encoder_type: 'FS2TransformerEncoder'
+    conf:
+      n_layers: 4
+      n_heads: 2
+      hidden_dim: 256
+      dropout: 0.25
+      d_inner: 1024
+      max_len: 2048
+
+decoder:
+    decoder_type: 'FS2TransformerDecoder'
+    input_dim: 256 # should be the same as the output of encoder
+    n_layers: 4
+    n_heads: 2
+    hidden_dim: 256
+    d_inner: 1024
+    dropout: 0.25
+    max_len: 2048 # max len of seq, for position embedding pre-computation
+
+#(class) Decoder(input_dim: int = 256, n_layers: int = 4, n_heads: int = 2,
+#hidden_dim: int = 256, d_inner: int = 1024, dropout: float = 0.5, max_len: int = 2048
+postnet:
+  postnet_type: 'PostUNet' # 'PostUNet', 'PostNet1d'
+speaker_embedding:
+    enable: True
+    vocab: #None
+    vocab_size: 1 #
+    weight: 1.0
+    dim: 256
+
+utterence_embedding:
+    enable: False
+    type: 'lstm' # resnet
+    feature_config:
+      type: 'mel'
+      n_mels: 80
+      sampling_rate: 22050
+      n_fft: 1024
+      hop_length: 256
+      win_length: 1024
+
+    model_config:
+      n_layers: 3
+      bidirectional: True
+
+hanzi_embedding:
+    enable: True
+    type: embedding
+    vocab: './gp.vocab'
+    dim: 256
+    weight: 0.5
+    max_seq_len: 100
+
+pinyin_embedding:
+    enable: True
+    type: embedding
+    vocab: './py.vocab'
+    dim: 256
+    weight: 1.0
+    max_seq_len: 100
+
+duration_predictor:
+  input_dim: 256 # should be the same as encoder hiddien_dim
+  filter_size: 256
+  kernel_size: 3
+  dropout: 0.5
+  duration_mean: 21.517294924096635 #for aishell3
+
+f0_predictor:
+  enable: False
+  filter_size: 256
+  kernel_size: 3
+  dropout: 0.5
+  n_bins: 256
+synthesis:
+  normalize: True # normalize the sound volume

examples/biaobei/gp.vocab

@@ -0,0 +1,4131 @@
+啊
+阿
+哀
+埃
+癌
+捱
+矮
+碍
+霭
+爱
+艾
+挨
+隘
+蔼
+暧
+哎
+黯
+安
+胺
+俺
+铵
+岸
+氨
+按
+鞍
+案
+唵
+暗
+昂
+盎
+肮
+熬
+澳
+凹
+翱
+敖
+坳
+袄
+懊
+奥
+傲
+獒
+拗
+笆
+靶
+灞
+拔
+跋
+芭
+峇
+爸
+吧
+巴
+疤
+叭
+扒
+罢
+坝
+粑
+霸
+八
+把
+败
+擘
+白
+摆
+百
+拜
+掰
+柏
+班
+瓣
+扮
+伴
+拌
+扳
+斑
+坂
+版
+颁
+般
+半
+搬
+板
+办
+癍
+镑
+榜
+棒
+谤
+磅
+膀
+邦
+蚌
+傍
+帮
+绑
+包
+葆
+煲
+宝
+胞
+薄
+饱
+堡
+报
+豹
+雹
+爆
+苞
+鲍
+抱
+暴
+保
+惫
+蓓
+狈
+杯
+倍
+背
+呗
+北
+卑
+悲
+碑
+辈
+备
+贝
+被
+犇
+苯
+笨
+奔
+本
+崩
+泵
+蹦
+绷
+甭
+嘣
+迸
+碧
+必
+逼
+臂
+痹
+璧
+毕
+壁
+弊
+鼻
+蔽
+笔
+比
+避
+闭
+彼
+币
+毙
+贬
+辩
+蝙
+编
+辨
+边
+鞭
+扁
+变
+匾
+遍
+便
+标
+镖
+表
+俵
+彪
+飙
+裱
+鳔
+摽
+别
+瘪
+鳖
+憋
+滨
+斌
+宾
+彬
+殡
+缤
+鬓
+槟
+饼
+兵
+柄
+秉
+并
+丙
+炳
+病
+冰
+礴
+饽
+跛
+搏
+舶
+帛
+伯
+脖
+卜
+驳
+拨
+膊
+勃
+箔
+播
+簸
+渤
+剥
+亳
+博
+波
+菠
+玻
+簿
+瓿
+部
+埠
+不
+补
+布
+捕
+哺
+步
+怖
+嚓
+擦
+礤
+彩
+睬
+材
+猜
+财
+裁
+采
+才
+踩
+偲
+菜
+寀
+蔡
+餐
+残
+蚕
+掺
+惭
+孱
+参
+灿
+璨
+惨
+藏
+苍
+仓
+舱
+沧
+槽
+嘈
+操
+草
+曹
+糙
+册
+策
+测
+厕
+侧
+涔
+岑
+蹭
+曾
+噌
+层
+差
+茶
+诧
+衩
+插
+察
+查
+岔
+叉
+杈
+拆
+豺
+钗
+柴
+馋
+蝉
+缠
+忏
+搀
+禅
+谄
+阐
+产
+浐
+铲
+颤
+肠
+场
+唱
+畅
+嫦
+常
+猖
+昌
+敞
+尝
+偿
+娼
+倡
+厂
+巢
+超
+吵
+嘲
+钞
+潮
+炒
+抄
+朝
+掣
+撤
+彻
+车
+扯
+澈
+衬
+琛
+郴
+嗔
+碜
+尘
+臣
+晨
+陈
+趁
+辰
+谌
+沉
+骋
+称
+惩
+成
+呈
+丞
+晟
+程
+橙
+诚
+澄
+撑
+乘
+秤
+承
+城
+逞
+瞠
+尺
+斥
+齿
+驰
+螭
+敕
+翅
+耻
+炽
+迟
+持
+叱
+池
+侈
+赤
+吃
+痴
+崇
+冲
+宠
+充
+虫
+仇
+踌
+绸
+愁
+筹
+臭
+瞅
+酬
+丑
+稠
+抽
+搐
+橱
+躇
+处
+除
+储
+初
+础
+锄
+厨
+畜
+楚
+雏
+出
+触
+揣
+啜
+踹
+串
+穿
+舛
+川
+喘
+传
+船
+床
+窗
+闯
+怆
+幢
+疮
+创
+吹
+椎
+锤
+炊
+槌
+垂
+捶
+唇
+蠢
+纯
+春
+醇
+淳
+绰
+辍
+龊
+戳
+糍
+词
+疵
+此
+呲
+磁
+次
+赐
+刺
+祠
+辞
+瓷
+慈
+雌
+茨
+从
+匆
+囱
+丛
+淙
+葱
+聪
+凑
+促
+殂
+醋
+簇
+粗
+猝
+篡
+撺
+蹿
+窜
+崔
+萃
+催
+脆
+悴
+翠
+摧
+璀
+粹
+存
+忖
+村
+寸
+搓
+磋
+矬
+嵯
+撮
+蹉
+错
+措
+挫
+厝
+痤
+靼
+答
+哒
+鞑
+达
+大
+跶
+瘩
+搭
+沓
+打
+褡
+耷
+代
+带
+黛
+贷
+袋
+待
+逮
+傣
+呆
+戴
+歹
+怠
+丹
+耽
+旦
+淡
+胆
+弹
+但
+殚
+单
+担
+郸
+蛋
+诞
+挡
+荡
+党
+当
+档
+铛
+道
+倒
+刀
+稻
+到
+导
+捣
+盗
+叨
+岛
+悼
+蹈
+祷
+得
+德
+嘚
+的
+蹬
+瞪
+邓
+灯
+凳
+等
+登
+地
+狄
+低
+诋
+敌
+翟
+笛
+缔
+抵
+滴
+递
+弟
+蒂
+第
+堤
+帝
+荻
+迪
+底
+癫
+电
+典
+巅
+殿
+淀
+碘
+颠
+滇
+甸
+掂
+奠
+惦
+点
+垫
+店
+钓
+雕
+凋
+屌
+调
+貂
+叼
+掉
+吊
+刁
+碟
+嗲
+牒
+跌
+耋
+叠
+蝶
+哋
+爹
+谍
+迭
+丁
+顶
+盯
+鼎
+酊
+叮
+铤
+定
+钉
+订
+丢
+冬
+峒
+洞
+栋
+董
+动
+冻
+懂
+东
+痘
+陡
+斗
+抖
+兜
+都
+窦
+逗
+豆
+杜
+堵
+犊
+毒
+赌
+妒
+渡
+镀
+督
+度
+嘟
+睹
+读
+肚
+独
+端
+断
+短
+锻
+段
+堆
+对
+队
+兑
+墩
+盾
+趸
+敦
+炖
+盹
+顿
+沌
+蹲
+囤
+吨
+躲
+多
+哆
+惰
+夺
+掇
+堕
+垛
+剁
+跺
+舵
+朵
+踱
+颚
+额
+鳄
+峨
+婀
+扼
+俄
+鄂
+愕
+呃
+娥
+遏
+恶
+饿
+厄
+讹
+鹅
+诶
+恩
+摁
+尔
+而
+洱
+二
+儿
+饵
+耳
+伐
+发
+珐
+法
+乏
+罚
+阀
+翻
+反
+番
+饭
+樊
+梵
+范
+犯
+返
+烦
+繁
+帆
+贩
+泛
+凡
+肪
+坊
+纺
+方
+仿
+放
+妨
+房
+访
+芳
+舫
+防
+妃
+菲
+非
+废
+诽
+斐
+飞
+绯
+啡
+费
+肺
+肥
+霏
+狒
+吠
+匪
+分
+坟
+芬
+粉
+氛
+吩
+粪
+份
+焚
+汾
+愤
+奋
+纷
+奉
+峰
+讽
+沣
+风
+丰
+缝
+锋
+蜂
+凤
+封
+逢
+冯
+疯
+枫
+否
+辅
+孵
+负
+斧
+福
+佛
+赋
+符
+府
+浮
+副
+腹
+阜
+袱
+服
+覆
+俯
+夫
+腐
+富
+幅
+敷
+妇
+伏
+肤
+附
+辐
+复
+俘
+蝠
+抚
+赴
+付
+芙
+父
+扶
+涪
+拂
+甫
+傅
+尬
+伽
+尕
+噶
+嘎
+旮
+垓
+改
+盖
+钙
+溉
+该
+概
+丐
+赶
+柑
+肝
+秆
+竿
+甘
+赣
+干
+敢
+尴
+感
+杆
+冈
+岗
+缸
+刚
+杠
+港
+钢
+糕
+膏
+羔
+高
+搞
+告
+稿
+睾
+歌
+革
+格
+阁
+戈
+割
+咯
+个
+骼
+镉
+鸽
+嗝
+哥
+胳
+搁
+铬
+葛
+疙
+隔
+各
+给
+亘
+艮
+跟
+根
+埂
+耕
+庚
+羹
+梗
+更
+哽
+耿
+蚣
+共
+贡
+公
+宫
+龚
+供
+恭
+拱
+弓
+攻
+肱
+巩
+工
+功
+构
+购
+够
+苟
+沟
+勾
+垢
+狗
+钩
+枸
+锢
+姑
+咕
+估
+谷
+鼓
+菇
+雇
+顾
+固
+孤
+古
+故
+毂
+呱
+骨
+鸪
+股
+寡
+刮
+卦
+剐
+挂
+瓜
+乖
+拐
+怪
+莞
+馆
+棺
+关
+管
+冠
+官
+灌
+贯
+罐
+观
+惯
+逛
+光
+广
+胱
+跪
+瑰
+龟
+傀
+硅
+桂
+规
+贵
+诡
+归
+鬼
+轨
+癸
+闺
+柜
+滚
+棍
+过
+国
+锅
+果
+郭
+椁
+裹
+蛤
+哈
+骸
+害
+还
+嗨
+孩
+亥
+海
+悍
+憨
+旱
+汉
+涵
+罕
+翰
+含
+喊
+撼
+酣
+阚
+韩
+捍
+鼾
+瀚
+寒
+汗
+邯
+憾
+蚶
+焊
+函
+杭
+夯
+沆
+航
+毫
+嚎
+皓
+好
+浩
+壕
+豪
+郝
+蚝
+号
+濠
+耗
+昊
+合
+禾
+喝
+颌
+盒
+褐
+赫
+核
+呵
+鹤
+何
+河
+荷
+贺
+和
+嘿
+黑
+恨
+很
+狠
+痕
+哼
+衡
+亨
+恒
+横
+轰
+哄
+吽
+讧
+宏
+虹
+洪
+鸿
+弘
+烘
+红
+泓
+吼
+候
+喉
+后
+侯
+猴
+厚
+狐
+呼
+湖
+户
+乎
+瑚
+鹕
+葫
+惚
+沪
+虎
+互
+囫
+唬
+胡
+忽
+护
+糊
+蝴
+壶
+化
+猾
+滑
+哗
+华
+划
+花
+嬅
+画
+话
+坏
+淮
+怀
+徊
+槐
+踝
+痪
+桓
+患
+焕
+唤
+缓
+涣
+换
+环
+幻
+欢
+浣
+蝗
+湟
+滉
+簧
+凰
+谎
+徨
+皇
+黄
+幌
+潢
+恍
+慌
+熀
+煌
+荒
+磺
+晃
+晖
+讳
+悔
+恢
+惠
+辉
+洄
+汇
+灰
+蕙
+晦
+秽
+翚
+贿
+珲
+绘
+回
+会
+慧
+挥
+徽
+毁
+婚
+魂
+混
+馄
+昏
+浑
+伙
+或
+火
+货
+惑
+豁
+获
+霍
+活
+祸
+基
+记
+伎
+继
+绩
+计
+忌
+肌
+籍
+挤
+圾
+既
+际
+急
+辑
+饥
+冀
+极
+骥
+迹
+姬
+缉
+祭
+季
+寂
+寄
+集
+鸡
+藉
+汲
+脊
+机
+岌
+几
+暨
+髻
+积
+嫉
+击
+戟
+疾
+鲫
+及
+级
+犄
+箕
+技
+纪
+剂
+己
+妓
+棘
+即
+济
+吉
+激
+唧
+讥
+家
+架
+驾
+颊
+夹
+假
+佳
+加
+价
+贾
+嘉
+钾
+甲
+茄
+戛
+稼
+迦
+嫁
+箭
+剪
+键
+贱
+健
+锏
+荐
+俭
+拣
+简
+尖
+肩
+鉴
+件
+检
+柬
+奸
+涧
+见
+艰
+謇
+囝
+舰
+捡
+坚
+剑
+毽
+监
+建
+减
+间
+煎
+兼
+渐
+践
+疆
+绛
+蒋
+匠
+缰
+将
+奖
+僵
+浆
+酱
+桨
+江
+讲
+降
+姜
+绞
+叫
+狡
+脚
+骄
+缴
+椒
+娇
+郊
+皎
+跤
+交
+胶
+浇
+矫
+剿
+礁
+教
+佼
+蕉
+焦
+角
+铰
+搅
+侥
+饺
+较
+酵
+轿
+蛟
+婕
+皆
+孑
+姐
+洁
+解
+介
+戒
+捷
+节
+阶
+竭
+截
+睫
+街
+借
+秸
+杰
+诫
+届
+劫
+揭
+嗟
+结
+接
+界
+进
+津
+谨
+仅
+禁
+靳
+瑾
+浸
+尽
+缙
+金
+近
+晋
+劲
+襟
+巾
+今
+锦
+筋
+紧
+斤
+镜
+井
+睛
+竟
+静
+竞
+茎
+晶
+靖
+痉
+颈
+鲸
+警
+景
+精
+经
+荆
+径
+惊
+敬
+净
+璟
+境
+京
+炯
+窘
+迥
+酒
+疚
+臼
+久
+究
+纠
+咎
+旧
+舅
+九
+柩
+救
+就
+揪
+韭
+聚
+居
+飓
+掬
+虡
+沮
+具
+炬
+举
+惧
+桔
+拘
+拒
+局
+巨
+距
+菊
+句
+剧
+橘
+据
+趄
+狙
+俱
+矩
+卷
+眷
+捐
+娟
+厥
+诀
+噘
+嚼
+掘
+撅
+绝
+决
+觉
+倔
+浚
+骏
+竣
+菌
+均
+俊
+君
+郡
+峻
+军
+咔
+咖
+喀
+卡
+凯
+开
+慨
+恺
+忾
+楷
+嘅
+砍
+侃
+刊
+槛
+看
+坎
+瞰
+堪
+勘
+康
+慷
+抗
+扛
+亢
+烤
+犒
+考
+铐
+尻
+靠
+客
+棵
+磕
+颗
+柯
+科
+课
+可
+苛
+刻
+嗑
+克
+壳
+瞌
+渴
+咳
+坷
+肯
+啃
+恳
+垦
+吭
+铿
+坑
+孔
+恐
+空
+控
+寇
+叩
+扣
+口
+抠
+枯
+酷
+裤
+刳
+库
+哭
+苦
+窟
+垮
+夸
+跨
+块
+快
+款
+宽
+况
+匡
+狂
+筐
+框
+矿
+哐
+眶
+骙
+煃
+蝰
+葵
+窥
+跬
+睽
+溃
+馈
+揆
+匮
+盔
+亏
+逵
+愧
+魁
+奎
+焜
+困
+坤
+昆
+捆
+括
+廓
+阔
+扩
+喇
+腊
+垃
+蜡
+邋
+剌
+辣
+拉
+啦
+旯
+赖
+睐
+来
+籁
+莱
+拦
+览
+懒
+烂
+滥
+澜
+缆
+揽
+篮
+栏
+兰
+榄
+蓝
+岚
+琅
+廊
+郎
+塱
+榔
+浪
+狼
+朗
+啷
+佬
+涝
+烙
+姥
+牢
+劳
+老
+捞
+潦
+肋
+乐
+了
+类
+蕾
+磊
+儡
+累
+泪
+垒
+勒
+雷
+嘞
+羸
+愣
+冷
+崚
+棱
+楞
+塄
+罹
+隶
+郦
+篱
+立
+粒
+理
+娌
+莉
+黎
+狸
+鹂
+厘
+俐
+漓
+璃
+砾
+力
+丽
+喱
+李
+离
+藜
+里
+励
+梨
+礼
+历
+俪
+例
+利
+呖
+吏
+鲤
+厉
+哩
+荔
+俩
+裢
+琏
+敛
+链
+帘
+莲
+连
+涟
+脸
+练
+联
+怜
+恋
+廉
+炼
+粮
+良
+踉
+量
+晾
+粱
+凉
+辆
+两
+梁
+谅
+亮
+寥
+缭
+廖
+聊
+疗
+撩
+料
+僚
+辽
+嘹
+趔
+裂
+咧
+猎
+冽
+烈
+列
+劣
+鬣
+赁
+吝
+林
+鳞
+淋
+凛
+拎
+临
+琳
+霖
+檩
+邻
+廪
+蔺
+麟
+零
+龄
+伶
+陵
+玲
+聆
+领
+另
+灵
+令
+菱
+羚
+祾
+姈
+铃
+凌
+岭
+柳
+瘤
+榴
+绺
+浏
+流
+留
+琉
+六
+遛
+刘
+硫
+溜
+茏
+笼
+龙
+陇
+聋
+咙
+胧
+窿
+隆
+拢
+垄
+蝼
+娄
+楼
+搂
+镂
+瘘
+篓
+喽
+漏
+陋
+掳
+鹿
+路
+碌
+炉
+逯
+卢
+陆
+鹭
+噜
+芦
+禄
+卤
+庐
+颅
+赂
+鲁
+露
+录
+撸
+孪
+滦
+卵
+銮
+挛
+栾
+峦
+乱
+囵
+沦
+抡
+伦
+轮
+仑
+论
+洛
+锣
+珞
+骡
+逻
+罗
+落
+螺
+络
+裸
+萝
+啰
+骆
+铝
+绿
+滤
+履
+驴
+吕
+捋
+旅
+律
+率
+侣
+虑
+屡
+略
+掠
+妈
+嘛
+麻
+骂
+玛
+马
+蚂
+码
+孖
+吗
+蟆
+脉
+麦
+迈
+埋
+卖
+买
+霾
+蛮
+鳗
+慢
+瞒
+馒
+曼
+满
+嫚
+蔓
+漫
+忙
+氓
+盲
+茫
+莽
+芒
+蟒
+耄
+毛
+貌
+冒
+茂
+贸
+冇
+矛
+卯
+髦
+峁
+锚
+铆
+茅
+帽
+茆
+猫
+泖
+么
+霉
+枚
+媚
+眉
+玫
+媒
+酶
+煤
+妹
+昧
+没
+梅
+魅
+每
+美
+寐
+莓
+们
+扪
+闷
+门
+蜢
+孟
+蒙
+盟
+猛
+勐
+懵
+萌
+梦
+朦
+锰
+泌
+迷
+觅
+咪
+洣
+蜜
+秘
+弥
+幂
+眯
+靡
+米
+密
+缅
+面
+免
+绵
+勉
+棉
+腼
+眠
+娩
+藐
+苗
+瞄
+渺
+喵
+妙
+庙
+描
+缈
+淼
+秒
+蔑
+乜
+灭
+咩
+悯
+敏
+抿
+泯
+民
+闽
+鸣
+冥
+命
+铭
+明
+茗
+名
+酩
+谬
+沫
+默
+摸
+膜
+抹
+摩
+蘑
+寞
+模
+漠
+末
+陌
+魔
+墨
+莫
+馍
+摹
+磨
+某
+缪
+谋
+幕
+姆
+牧
+木
+睦
+穆
+亩
+募
+墓
+拇
+慕
+母
+目
+嗯
+钠
+拿
+哪
+娜
+呐
+那
+纳
+捺
+吶
+肭
+孻
+乃
+奈
+奶
+耐
+南
+男
+囡
+赧
+腩
+难
+湳
+楠
+蝻
+囊
+囔
+馕
+垴
+挠
+闹
+恼
+脑
+婥
+孬
+讷
+呢
+馁
+内
+恁
+嫩
+能
+昵
+逆
+妮
+拟
+尼
+你
+霓
+倪
+泥
+匿
+腻
+溺
+年
+辗
+碾
+撵
+捻
+黏
+鲶
+蔫
+念
+拈
+酿
+娘
+嬢
+尿
+鸟
+袅
+涅
+孽
+聂
+捏
+镊
+蹑
+您
+泞
+咛
+凝
+宁
+拧
+妞
+纽
+扭
+牛
+弄
+哝
+浓
+农
+奴
+弩
+怒
+努
+暖
+糯
+诺
+挪
+懦
+女
+虐
+喔
+噢
+哦
+偶
+怄
+欧
+呕
+殴
+沤
+鸥
+趴
+杷
+爬
+葩
+琶
+啪
+怕
+帕
+徘
+牌
+湃
+拍
+派
+排
+攀
+畔
+潘
+判
+盼
+盘
+泮
+叛
+彷
+胖
+乓
+旁
+螃
+庞
+滂
+泡
+咆
+狍
+袍
+抛
+炮
+跑
+刨
+裴
+配
+沛
+培
+佩
+赔
+胚
+陪
+喷
+盆
+鹏
+硼
+彭
+澎
+碰
+烹
+怦
+嘭
+蓬
+膨
+篷
+捧
+抨
+砰
+朋
+棚
+披
+譬
+纰
+毗
+媲
+坯
+痞
+癖
+皮
+劈
+辟
+屁
+匹
+批
+蜱
+疲
+噼
+脾
+啤
+枇
+僻
+琵
+翩
+片
+偏
+骗
+篇
+嫖
+剽
+瓢
+漂
+缥
+瞟
+飘
+票
+撇
+瞥
+品
+拼
+频
+聘
+贫
+屏
+平
+坪
+苹
+乒
+萍
+凭
+瓶
+评
+叵
+颇
+泼
+鄱
+迫
+珀
+婆
+破
+坡
+泊
+魄
+剖
+埔
+普
+葡
+铺
+蹼
+浦
+溥
+莆
+菩
+仆
+曝
+朴
+蒲
+谱
+圃
+濮
+扑
+凄
+麒
+器
+岐
+綮
+崎
+七
+戚
+脐
+琦
+琪
+祈
+泣
+弃
+亓
+砌
+漆
+杞
+婍
+绮
+期
+汽
+棋
+淇
+契
+奇
+歧
+其
+妻
+启
+栖
+旗
+圻
+玘
+齐
+鳍
+祁
+欺
+骐
+骑
+起
+企
+迄
+乞
+岂
+气
+洽
+掐
+恰
+虔
+前
+签
+堑
+倩
+浅
+牵
+潜
+嵌
+迁
+铅
+千
+钱
+茜
+綪
+欠
+遣
+钳
+谦
+乾
+黔
+谴
+歉
+锵
+强
+抢
+蔷
+呛
+墙
+羌
+腔
+跄
+枪
+戗
+桥
+鞘
+巧
+悄
+憔
+俏
+乔
+侨
+峭
+翘
+瞧
+敲
+窍
+撬
+锹
+惬
+窃
+怯
+切
+妾
+且
+亲
+揿
+琴
+沁
+擒
+禽
+芹
+寝
+勤
+嗪
+侵
+钦
+秦
+晴
+蜻
+倾
+清
+情
+磬
+顷
+庆
+请
+卿
+氰
+轻
+擎
+氢
+青
+筇
+穷
+穹
+琼
+銶
+逑
+糗
+酋
+秋
+丘
+求
+邱
+球
+鳅
+囚
+裘
+躯
+取
+趋
+去
+曲
+区
+屈
+趣
+岖
+驱
+渠
+黢
+觑
+瞿
+娶
+全
+蜷
+犬
+痊
+权
+券
+颧
+醛
+劝
+泉
+圈
+拳
+确
+鹊
+缺
+雀
+却
+瘸
+裙
+逡
+群
+染
+然
+燃
+苒
+冉
+嚷
+攘
+瓤
+让
+壤
+穰
+绕
+饶
+扰
+娆
+惹
+热
+仞
+任
+人
+韧
+刃
+认
+饪
+仁
+忍
+纫
+荏
+扔
+仍
+日
+冗
+茸
+荣
+融
+绒
+镕
+容
+蓉
+溶
+柔
+肉
+揉
+媃
+褥
+辱
+乳
+濡
+汝
+如
+儒
+茹
+入
+软
+阮
+睿
+蕤
+锐
+芮
+蕊
+瑞
+润
+若
+弱
+仨
+洒
+萨
+撒
+噻
+腮
+塞
+赛
+馓
+三
+散
+伞
+丧
+桑
+嗓
+搡
+扫
+嫂
+瘙
+臊
+骚
+瑟
+啬
+涩
+色
+森
+僧
+刹
+砂
+杀
+莎
+厦
+傻
+沙
+煞
+鲨
+啥
+霎
+纱
+晒
+筛
+善
+衫
+擅
+缮
+珊
+膳
+删
+蟮
+山
+闪
+鳝
+扇
+讪
+陕
+赡
+伤
+赏
+晌
+商
+尚
+裳
+上
+劭
+韶
+稍
+哨
+捎
+少
+绍
+勺
+邵
+芍
+潲
+梢
+烧
+社
+舌
+麝
+奢
+蛇
+摄
+射
+歙
+涉
+设
+舍
+谁
+慎
+深
+申
+呻
+绅
+沈
+身
+婶
+肾
+审
+神
+渗
+甚
+什
+伸
+声
+剩
+牲
+圣
+盛
+绳
+甥
+省
+笙
+胜
+生
+升
+蚀
+湿
+尸
+奭
+辻
+逝
+栻
+虱
+恃
+失
+驶
+识
+氏
+视
+示
+匙
+是
+世
+狮
+释
+嗜
+诗
+拾
+似
+式
+势
+仕
+事
+士
+适
+使
+舐
+噬
+柿
+时
+屎
+食
+寔
+拭
+始
+室
+试
+实
+施
+石
+誓
+史
+市
+饰
+十
+矢
+师
+守
+首
+受
+收
+授
+售
+瘦
+手
+兽
+寿
+纾
+淑
+数
+蜀
+漱
+输
+属
+署
+鼠
+殊
+舒
+竖
+孰
+薯
+抒
+书
+赎
+蔬
+束
+叔
+树
+枢
+梳
+熟
+疏
+暑
+墅
+术
+黍
+述
+刷
+耍
+帅
+衰
+蟀
+摔
+甩
+涮
+栓
+拴
+双
+霜
+爽
+水
+税
+睡
+瞬
+顺
+吮
+朔
+说
+烁
+硕
+死
+寺
+嗣
+四
+肆
+嘶
+祀
+厮
+泗
+巳
+蛳
+私
+斯
+司
+饲
+撕
+思
+丝
+sil
+怂
+讼
+悚
+耸
+嵩
+宋
+松
+送
+叟
+擞
+嗽
+馊
+艘
+薮
+搜
+sp1
+稣
+速
+苏
+诉
+俗
+塑
+肃
+宿
+溯
+粟
+素
+夙
+算
+蒜
+酸
+碎
+隧
+绥
+岁
+遂
+髓
+濉
+随
+虽
+隋
+穗
+邃
+荪
+榫
+损
+孙
+唆
+索
+唢
+所
+琐
+锁
+缩
+嗦
+榻
+她
+蹋
+塔
+它
+他
+踏
+塌
+遢
+挞
+台
+泰
+态
+肽
+抬
+太
+胎
+枱
+汰
+坦
+叹
+昙
+谈
+贪
+潭
+碳
+覃
+毯
+袒
+摊
+坍
+痰
+炭
+滩
+探
+坛
+谭
+忐
+瘫
+搪
+棠
+唐
+躺
+汤
+烫
+倘
+糖
+蹚
+趟
+塘
+螳
+淌
+堂
+膛
+套
+掏
+饕
+淘
+滔
+逃
+桃
+陶
+涛
+萄
+讨
+韬
+特
+忒
+忑
+滕
+疼
+誊
+腾
+藤
+踢
+体
+替
+啼
+梯
+屉
+惕
+蹄
+缇
+涕
+剔
+鹈
+剃
+提
+题
+腆
+填
+添
+甜
+田
+天
+恬
+舔
+笤
+眺
+窕
+条
+挑
+跳
+铁
+餮
+帖
+贴
+听
+亭
+艇
+挺
+庭
+霆
+厅
+蜓
+廷
+停
+婷
+痛
+佟
+铜
+桶
+通
+童
+同
+统
+筒
+桐
+捅
+彤
+投
+偷
+头
+透
+吐
+徒
+屠
+图
+土
+秃
+途
+涂
+兔
+凸
+突
+湍
+团
+疃
+彖
+退
+腿
+蜕
+颓
+推
+屯
+豚
+吞
+饨
+暾
+臀
+庹
+驮
+托
+妥
+跎
+鸵
+驼
+唾
+陀
+沱
+脱
+拖
+拓
+砣
+蛙
+袜
+娃
+哇
+腽
+挖
+瓦
+洼
+崴
+歪
+外
+菀
+腕
+蜿
+皖
+挽
+玩
+万
+丸
+完
+婉
+宛
+晚
+弯
+碗
+顽
+湾
+汪
+亡
+王
+网
+妄
+旺
+忘
+望
+罔
+往
+枉
+违
+伟
+卫
+为
+惟
+韦
+危
+畏
+喂
+娓
+蔚
+未
+位
+微
+帷
+魏
+巍
+纬
+圩
+嵬
+渭
+伪
+维
+炜
+委
+唯
+胃
+尾
+诿
+慰
+围
+薇
+味
+萎
+猥
+苇
+威
+谓
+蚊
+汶
+文
+雯
+温
+纹
+闻
+紊
+吻
+问
+稳
+嗡
+翁
+蓊
+瓮
+我
+龌
+涡
+斡
+蜗
+沃
+握
+窝
+渥
+卧
+物
+侮
+雾
+五
+午
+乌
+蜈
+芜
+务
+伍
+毋
+呜
+武
+悟
+吾
+舞
+屋
+晤
+误
+鹉
+妩
+勿
+梧
+吴
+巫
+无
+捂
+诬
+唔
+污
+西
+犀
+细
+希
+熄
+禧
+熙
+蟋
+隙
+膝
+席
+唏
+戏
+洗
+袭
+悉
+昔
+惜
+晰
+兮
+嬉
+吸
+玺
+夕
+锡
+羲
+徙
+嘻
+系
+析
+牺
+皙
+息
+蜥
+稀
+习
+媳
+溪
+喜
+侠
+瞎
+瑕
+霞
+峡
+虾
+狭
+下
+辖
+暇
+吓
+夏
+显
+馅
+纤
+县
+咸
+现
+羡
+贤
+舷
+限
+献
+鲜
+先
+弦
+险
+腺
+掀
+线
+蚬
+陷
+睍
+嫌
+宪
+仙
+衔
+闲
+涎
+相
+项
+想
+像
+详
+享
+向
+襄
+箱
+巷
+镶
+乡
+橡
+湘
+祥
+飨
+香
+厢
+翔
+响
+象
+销
+淆
+孝
+硝
+嚣
+筱
+小
+霄
+笑
+消
+逍
+哮
+啸
+潇
+肖
+晓
+萧
+校
+效
+懈
+瀣
+械
+些
+屑
+胁
+写
+歇
+偕
+协
+挟
+燮
+谢
+携
+亵
+卸
+邪
+谐
+鞋
+泄
+蟹
+泻
+斜
+新
+心
+芯
+馨
+衅
+辛
+鑫
+炘
+昕
+薪
+欣
+信
+姓
+行
+刑
+型
+邢
+陉
+星
+兴
+形
+腥
+幸
+性
+猩
+醒
+杏
+胸
+熊
+匈
+凶
+雄
+兄
+汹
+休
+修
+咻
+锈
+袖
+绣
+秀
+羞
+朽
+许
+徐
+序
+绪
+墟
+旭
+顼
+煦
+胥
+虚
+吁
+续
+嘘
+絮
+栩
+婿
+须
+诩
+叙
+需
+蓄
+宣
+喧
+癣
+旋
+瑄
+轩
+漩
+悬
+玄
+选
+渲
+暄
+绚
+炫
+鳕
+靴
+削
+雪
+薛
+血
+穴
+学
+训
+讯
+汛
+逊
+寻
+询
+旬
+迅
+勋
+窨
+循
+驯
+熏
+巡
+押
+雅
+哑
+亚
+涯
+崖
+鸦
+丫
+鸭
+娅
+衙
+轧
+芽
+呀
+讶
+压
+桠
+吖
+牙
+嫣
+焱
+验
+延
+琰
+烟
+蜒
+盐
+岩
+淹
+闫
+颜
+炎
+砚
+檐
+堰
+演
+唁
+艳
+彦
+妍
+严
+研
+掩
+晏
+沿
+焰
+眼
+咽
+言
+雁
+厌
+偃
+湮
+俨
+嬿
+宴
+燕
+洋
+阳
+疡
+央
+垟
+羊
+秧
+泱
+扬
+氧
+殃
+样
+旸
+养
+佯
+仰
+漾
+杨
+恙
+痒
+肴
+夭
+杳
+窑
+遥
+榣
+谣
+窈
+耀
+钥
+瑶
+滧
+舀
+邀
+尧
+妖
+幺
+咬
+垚
+药
+吆
+摇
+腰
+姚
+要
+叶
+野
+揶
+业
+冶
+晔
+爷
+耶
+烨
+曳
+页
+噎
+也
+嘢
+液
+夜
+屹
+宜
+伊
+已
+疑
+蚁
+迤
+易
+蜴
+弋
+亿
+溢
+奕
+倚
+毅
+癔
+医
+仪
+逸
+翌
+绎
+颐
+疫
+翼
+夷
+弈
+一
+以
+益
+怡
+艺
+意
+轶
+姨
+贻
+祎
+议
+役
+薏
+乙
+咦
+谊
+忆
+移
+遗
+亦
+沂
+抑
+裔
+胰
+义
+译
+矣
+依
+椅
+彝
+壹
+痍
+异
+衣
+阴
+寅
+铟
+饮
+因
+引
+鄞
+瘾
+吟
+尹
+殷
+荫
+隐
+音
+姻
+印
+淫
+银
+茵
+垠
+颖
+莹
+盈
+英
+溁
+影
+蝇
+莺
+璎
+荧
+罂
+樱
+映
+赢
+楹
+迎
+婴
+硬
+营
+鹦
+萦
+应
+鹰
+喲
+哟
+唷
+涌
+用
+雍
+拥
+踊
+泳
+佣
+邕
+庸
+勇
+壅
+俑
+永
+咏
+恿
+莠
+右
+釉
+黝
+忧
+由
+牖
+柚
+莜
+有
+优
+犹
+佑
+悠
+酉
+邮
+游
+诱
+油
+呦
+幽
+又
+幼
+友
+攸
+尤
+渔
+予
+榆
+舆
+玉
+语
+禹
+愉
+于
+郁
+腴
+浴
+与
+遇
+迂
+淤
+愚
+瑜
+逾
+育
+屿
+俞
+预
+御
+鱼
+峪
+虞
+裕
+瘀
+狱
+雨
+欲
+谀
+娱
+余
+誉
+愈
+瑀
+喻
+域
+渝
+寓
+煜
+豫
+驭
+宇
+禺
+钰
+臾
+谕
+羽
+揄
+妤
+妪
+元
+源
+缘
+圆
+羱
+援
+渊
+愿
+辕
+远
+猿
+院
+原
+苑
+员
+嫄
+媛
+怨
+冤
+袁
+园
+跃
+越
+粤
+约
+曰
+月
+阅
+悦
+钺
+岳
+昀
+孕
+韵
+蕴
+运
+酝
+允
+殒
+耘
+纭
+匀
+陨
+晕
+云
+赟
+咋
+匝
+咂
+杂
+砸
+载
+宰
+崽
+仔
+灾
+在
+再
+栽
+昝
+錾
+簪
+暂
+糌
+咱
+攒
+赞
+脏
+臧
+葬
+赃
+澡
+凿
+蚤
+造
+枣
+遭
+燥
+灶
+躁
+噪
+皂
+糟
+早
+择
+则
+泽
+仄
+责
+贼
+怎
+增
+赠
+咤
+蚱
+诈
+柞
+眨
+闸
+栅
+札
+喳
+渣
+乍
+炸
+榨
+扎
+砟
+吒
+债
+宅
+斋
+窄
+寨
+摘
+绽
+詹
+占
+盏
+展
+沾
+战
+毡
+崭
+斩
+瞻
+粘
+蘸
+站
+獐
+长
+璋
+账
+胀
+障
+杖
+彰
+幛
+仗
+张
+丈
+帐
+涨
+樟
+漳
+章
+掌
+召
+兆
+爪
+肇
+罩
+昭
+赵
+招
+找
+照
+啫
+浙
+辙
+着
+鹧
+者
+哲
+遮
+蔗
+这
+折
+祯
+贞
+振
+枕
+鸩
+疹
+甄
+帧
+侦
+诊
+震
+缜
+镇
+赈
+斟
+针
+珍
+真
+阵
+圳
+蒸
+整
+筝
+争
+证
+征
+挣
+睁
+峥
+铮
+郑
+症
+正
+拯
+政
+枝
+支
+侄
+峙
+痣
+肢
+值
+芝
+致
+炙
+职
+脂
+直
+蜘
+置
+殖
+吱
+之
+栉
+稚
+智
+趾
+织
+植
+址
+痔
+秩
+至
+旨
+只
+治
+制
+执
+止
+帜
+志
+滞
+纸
+指
+质
+汁
+挚
+知
+钟
+众
+冢
+肿
+忠
+仲
+终
+重
+踵
+种
+中
+衷
+皱
+胄
+咒
+妯
+洲
+周
+肘
+舟
+帚
+昼
+骤
+轴
+宙
+州
+粥
+诸
+驻
+柱
+炷
+筑
+祝
+猪
+竹
+瞩
+朱
+煮
+珠
+拄
+助
+株
+主
+著
+注
+蛛
+住
+逐
+烛
+洙
+嘱
+抓
+拽
+专
+篆
+撰
+赚
+砖
+转
+庄
+状
+撞
+壮
+桩
+妆
+装
+赘
+惴
+缀
+追
+锥
+坠
+谆
+准
+捉
+桌
+酌
+茁
+卓
+浊
+拙
+镯
+淄
+资
+兹
+姊
+姿
+梓
+紫
+咨
+渍
+訾
+自
+龇
+滋
+字
+子
+宗
+纵
+踪
+鬃
+综
+棕
+总
+奏
+揍
+走
+邹
+卒
+祖
+诅
+组
+族
+足
+阻
+租
+钻
+纂
+缵
+攥
+嘴
+最
+醉
+罪
+尊
+遵
+做
+座
+佐
+作
+左
+坐
+昨
+咗
+琢
+Ｂ

examples/biaobei/input.txt

@@ -0,0 +1,13 @@
+#name|py|gp|spk
+text1|sil ni3 qu4 zuo4 fan4 ba5 sil|sil 你 去 做 饭 吧 sil|0
+text2|sil ni3 qu4 zuo4 fan4 ba5 sil|sil 你 去 做 饭 吧 sil|0
+text3|sil you3 ge4 xiao3 hai2 jiao4 xiao3 du4 sil|sil 有　个 小 孩 叫 小 杜 sil|0
+text4|sil shang4 jie1 da3 cu4 you4 mai3 bu4 sil|sil 上 街 打 醋 又 买 布 sil|0
+text5|sil mai3 le5 bu4 da3 le5 cu4 sil|sil 买 了 布 打 了 醋 sil|0
+text6|sil hui2 tou2 kan4 jian4 ying1 zhua1 tu4 sil|sil 回 头 看 见 鹰 抓 兔 sil|0
+text7|sil fang4 xia4 bu4 ge1 xia4 cu4 sil|sil 放 下 布 搁 下 醋 sil|0
+text8|sil shang4 qian2 qu4 zhui1 ying1 he2 tu4 sil|sil 上 前 去 追 鹰 和 兔 sil|0
+text9|sil fei1 le5 ying1 pao3 le5 tu4 sil|sil 飞 了 鹰 跑 了 兔 sil|0
+text10|sil sa3 le5 cu4 shi1 le5 bu4 sil|sil 洒 了 醋 湿 了 布 sil|0
+text11|sil wei4 shi4 ying4 xin1 de5 wang3 luo4 chuan2 bo1 fang1 shi4 he2 du2 zhe3 yue4 du2 xi2 guan4 sil|sil 为 适 应 新 的 网 络 传 播 方 式 和 读 者 阅 读 习 惯 sil|0
+text12|sil zhe4 ge4 xiao3 hair2 r zhen1 dong3 shir4 sil|sil 这 个 小 孩 儿 真 懂 事 sil|0

examples/biaobei/py.vocab

@@ -0,0 +1,1608 @@
+a1
+a4
+a5
+ai1
+ai2
+ai3
+ai4
+ai5
+an1
+an2
+an3
+an4
+ang1
+ang2
+ang4
+ao1
+ao2
+ao3
+ao4
+ba1
+ba2
+ba3
+ba4
+ba5
+bai1
+bai2
+bai3
+bai4
+bai5
+ban1
+ban2
+ban3
+ban4
+bang1
+bang2
+bang3
+bang4
+bang5
+banr3
+bao1
+bao2
+bao3
+bao4
+bao5
+bei1
+bei2
+bei3
+bei4
+bei5
+ben1
+ben2
+ben3
+ben4
+ben5
+beng1
+beng2
+beng4
+bi1
+bi2
+bi3
+bi4
+bian1
+bian2
+bian3
+bian4
+bian5
+biao1
+biao2
+biao3
+biao4
+bie1
+bie2
+bie3
+bie4
+bin1
+bin4
+bing1
+bing2
+bing3
+bing4
+bo1
+bo2
+bo3
+bo4
+bo5
+bu2
+bu3
+bu4
+bu5
+ca1
+ca3
+cai1
+cai2
+cai3
+cai4
+cai5
+can1
+can2
+can3
+can4
+cang1
+cang2
+cao1
+cao2
+cao3
+cao5
+ce4
+cen1
+cen2
+ceng1
+ceng2
+ceng4
+cha1
+cha2
+cha3
+cha4
+cha5
+chai1
+chai2
+chan1
+chan2
+chan3
+chan4
+chan5
+chang1
+chang2
+chang3
+chang4
+chang5
+chao1
+chao2
+chao3
+chao5
+che1
+che2
+che3
+che4
+chen1
+chen2
+chen3
+chen4
+chen5
+cheng1
+cheng2
+cheng3
+cheng4
+cheng5
+chi1
+chi2
+chi3
+chi4
+chi5
+chong1
+chong2
+chong3
+chong4
+chou1
+chou2
+chou3
+chou4
+chou5
+chu1
+chu2
+chu3
+chu4
+chu5
+chuai1
+chuai3
+chuai4
+chuan1
+chuan2
+chuan3
+chuan4
+chuan5
+chuang1
+chuang2
+chuang3
+chuang4
+chuang5
+chui1
+chui2
+chui5
+chun1
+chun2
+chun3
+chuo1
+chuo4
+ci1
+ci2
+ci3
+ci4
+cir2
+cong1
+cong2
+cong5
+cou4
+cu1
+cu2
+cu4
+cuan1
+cuan2
+cuan4
+cui1
+cui2
+cui3
+cui4
+cui5
+cun1
+cun2
+cun3
+cun4
+cuo1
+cuo2
+cuo4
+da1
+da2
+da3
+da4
+da5
+dai1
+dai3
+dai4
+dai5
+dan1
+dan2
+dan3
+dan4
+dan5
+dang1
+dang2
+dang3
+dang4
+dao1
+dao2
+dao3
+dao4
+dao5
+daor4
+de2
+de4
+de5
+dei2
+dei3
+deng1
+deng2
+deng3
+deng4
+di1
+di2
+di3
+di4
+di5
+dia2
+dia3
+dian1
+dian2
+dian3
+dian4
+dianr2
+dianr3
+diao1
+diao3
+diao4
+diao5
+die1
+die2
+die4
+die5
+ding1
+ding2
+ding3
+ding4
+dir2
+dir3
+dir4
+diu1
+dong1
+dong2
+dong3
+dong4
+dong5
+dou1
+dou2
+dou3
+dou4
+dou5
+du1
+du2
+du3
+du4
+du5
+duan1
+duan2
+duan3
+duan4
+dui1
+dui4
+duir1
+duir4
+dun1
+dun3
+dun4
+dunr3
+duo1
+duo2
+duo3
+duo4
+duo5
+duor3
+e1
+e2
+e3
+e4
+e5
+ei5
+en1
+en4
+er2
+er3
+er4
+er5
+fa1
+fa2
+fa3
+fa4
+fa5
+fan1
+fan2
+fan3
+fan4
+fan5
+fang1
+fang2
+fang3
+fang4
+fang5
+fanr4
+far2
+fei1
+fei2
+fei3
+fei4
+fen1
+fen2
+fen3
+fen4
+fen5
+feng1
+feng2
+feng3
+feng4
+feng5
+fo2
+fou2
+fou3
+fu1
+fu2
+fu3
+fu4
+fu5
+fur4
+ga1
+ga2
+ga3
+ga4
+gai1
+gai2
+gai3
+gai4
+gair4
+gan1
+gan2
+gan3
+gan4
+gan5
+gang1
+gang2
+gang3
+gang4
+gang5
+gao1
+gao2
+gao3
+gao4
+ge1
+ge2
+ge3
+ge4
+ge5
+gei2
+gei3
+gen1
+gen4
+geng1
+geng2
+geng3
+geng4
+genr1
+ger2
+ger3
+ger4
+gong1
+gong2
+gong3
+gong4
+gong5
+gou1
+gou2
+gou3
+gou4
+gou5
+gu1
+gu2
+gu3
+gu4
+gu5
+gua1
+gua2
+gua3
+gua4
+gua5
+guai1
+guai2
+guai3
+guai4
+guan1
+guan2
+guan3
+guan4
+guang1
+guang2
+guang3
+guang4
+guang5
+gui1
+gui2
+gui3
+gui4
+gui5
+gun2
+gun3
+gun4
+gunr4
+guo1
+guo2
+guo3
+guo4
+guo5
+ha1
+ha2
+ha5
+hai1
+hai2
+hai3
+hai4
+hai5
+hair2
+han1
+han2
+han3
+han4
+han5
+hang1
+hang2
+hang4
+hanr4
+hao1
+hao2
+hao3
+hao4
+hao5
+he1
+he2
+he4
+he5
+hei1
+hen2
+hen3
+hen4
+heng1
+heng2
+heng4
+heng5
+her2
+hong1
+hong2
+hong3
+hong4
+hou2
+hou3
+hou4
+hou5
+hu1
+hu2
+hu3
+hu4
+hu5
+hua1
+hua2
+hua4
+hua5
+huai2
+huai4
+huan1
+huan2
+huan3
+huan4
+huan5
+huang1
+huang2
+huang3
+huang4
+huang5
+huanr1
+hui1
+hui2
+hui3
+hui4
+hui5
+huir3
+huir4
+hun1
+hun2
+hun4
+hun5
+huo1
+huo2
+huo3
+huo4
+huo5
+huor2
+huor3
+ji1
+ji2
+ji3
+ji4
+ji5
+jia1
+jia2
+jia3
+jia4
+jia5
+jian1
+jian2
+jian3
+jian4
+jian5
+jiang1
+jiang2
+jiang3
+jiang4
+jiang5
+jiao1
+jiao2
+jiao3
+jiao4
+jiao5
+jiar1
+jie1
+jie2
+jie3
+jie4
+jie5
+jin1
+jin2
+jin3
+jin4
+jin5
+jing1
+jing2
+jing3
+jing4
+jing5
+jingr3
+jinr4
+jiong2
+jiong3
+jir1
+jiu1
+jiu2
+jiu3
+jiu4
+jiu5
+ju1
+ju2
+ju3
+ju4
+ju5
+juan1
+juan2
+juan3
+juan4
+jue1
+jue2
+jue4
+jun1
+jun4
+jun5
+ka1
+ka2
+ka3
+kai1
+kai2
+kai3
+kai4
+kan1
+kan2
+kan3
+kan4
+kan5
+kang1
+kang2
+kang4
+kao1
+kao2
+kao3
+kao4
+ke1
+ke2
+ke3
+ke4
+ken2
+ken3
+keng1
+kong1
+kong2
+kong3
+kong4
+kou1
+kou2
+kou3
+kou4
+kou5
+ku1
+ku2
+ku3
+ku4
+kua1
+kua2
+kua3
+kua4
+kuai4
+kuai5
+kuair4
+kuan1
+kuan2
+kuan3
+kuang1
+kuang2
+kuang4
+kuang5
+kui1
+kui2
+kui3
+kui4
+kun1
+kun2
+kun3
+kun4
+kuo4
+la1
+la2
+la3
+la4
+la5
+lai2
+lai4
+lai5
+lan2
+lan3
+lan4
+lan5
+lang1
+lang2
+lang3
+lang4
+lang5
+lanr4
+lao1
+lao2
+lao3
+lao4
+lao5
+le4
+le5
+lei1
+lei2
+lei3
+lei4
+lei5
+leng2
+leng3
+leng4
+li2
+li3
+li4
+li5
+lia2
+lia3
+lian2
+lian3
+lian4
+lian5
+liang2
+liang3
+liang4
+liang5
+liao2
+liao3
+liao4
+liao5
+liar3
+lie1
+lie2
+lie3
+lie4
+lie5
+lin1
+lin2
+lin3
+lin4
+lin5
+ling2
+ling3
+ling4
+ling5
+lir3
+liu1
+liu2
+liu3
+liu4
+liu5
+lo5
+long2
+long3
+long4
+long5
+lou1
+lou2
+lou3
+lou4
+lou5
+lu1
+lu2
+lu3
+lu4
+lu5
+luan2
+luan3
+luan4
+lun1
+lun2
+lun4
+luo1
+luo2
+luo3
+luo4
+luo5
+lv2
+lv3
+lv4
+lv5
+lve4
+ma1
+ma2
+ma3
+ma4
+ma5
+mai2
+mai3
+mai4
+mai5
+man2
+man3
+man4
+mang2
+mang3
+mao1
+mao2
+mao3
+mao4
+mao5
+me1
+me5
+mei1
+mei2
+mei3
+mei4
+mei5
+men1
+men2
+men4
+men5
+meng1
+meng2
+meng3
+meng4
+menr2
+menr5
+mi1
+mi2
+mi3
+mi4
+mi5
+mian2
+mian3
+mian4
+mian5
+miao1
+miao2
+miao3
+miao4
+mie1
+mie4
+min2
+min3
+ming2
+ming3
+ming4
+ming5
+mingr2
+miu4
+mo1
+mo2
+mo3
+mo4
+mo5
+mou2
+mou3
+mou5
+mu2
+mu3
+mu4
+mu5
+na2
+na3
+na4
+na5
+nai2
+nai3
+nai4
+nai5
+nan1
+nan2
+nan3
+nan4
+nan5
+nang1
+nang2
+nang5
+nao1
+nao2
+nao3
+nao4
+nao5
+nar2
+nar3
+nar4
+ne2
+ne4
+ne5
+nei3
+nei4
+nen4
+nen5
+neng2
+ng1
+ni1
+ni2
+ni3
+ni4
+nian1
+nian2
+nian3
+nian4
+niang2
+niang4
+niang5
+niao2
+niao3
+niao4
+nie1
+nie4
+nin2
+ning2
+ning3
+ning4
+niu1
+niu2
+niu3
+niu4
+niu5
+niur1
+nong2
+nong4
+nong5
+nu2
+nu3
+nu4
+nuan2
+nuan3
+nuo2
+nuo4
+nv2
+nv3
+nv5
+nve4
+o1
+o4
+o5
+ou1
+ou2
+ou3
+ou4
+pa1
+pa2
+pa4
+pa5
+pai1
+pai2
+pai4
+pai5
+pan1
+pan2
+pan4
+pan5
+pang1
+pang2
+pang4
+pao1
+pao2
+pao3
+pao4
+pei1
+pei2
+pei4
+pen1
+pen2
+pen4
+peng1
+peng2
+peng3
+peng4
+peng5
+pi1
+pi2
+pi3
+pi4
+pi5
+pian1
+pian2
+pian4
+pian5
+piao1
+piao2
+piao3
+piao4
+pie1
+pie2
+pie3
+pin1
+pin2
+pin3
+pin4
+ping1
+ping2
+ping5
+pir2
+po1
+po2
+po3
+po4
+po5
+pou1
+pu1
+pu2
+pu3
+pu4
+qi1
+qi2
+qi3
+qi4
+qi5
+qia1
+qia2
+qia3
+qia4
+qian1
+qian2
+qian3
+qian4
+qian5
+qiang1
+qiang2
+qiang3
+qiang4
+qiang5
+qiao1
+qiao2
+qiao3
+qiao4
+qiao5
+qie1
+qie2
+qie3
+qie4
+qie5
+qin1
+qin2
+qin3
+qin4
+qin5
+qing1
+qing2
+qing3
+qing4
+qing5
+qiong2
+qir4
+qiu1
+qiu2
+qiu3
+qiu5
+qu1
+qu2
+qu3
+qu4
+qu5
+quan1
+quan2
+quan3
+quan4
+quan5
+que1
+que2
+que4
+qun1
+qun2
+r
+ran2
+ran3
+ran5
+rang1
+rang2
+rang3
+rang4
+rang5
+rao2
+rao3
+rao4
+re2
+re3
+re4
+ren2
+ren3
+ren4
+ren5
+reng1
+reng2
+renr4
+ri4
+rong2
+rong3
+rou2
+rou4
+rou5
+ru2
+ru3
+ru4
+ruan2
+ruan3
+rui2
+rui3
+rui4
+run4
+run5
+ruo4
+sa1
+sa2
+sa3
+sa4
+sa5
+sai1
+sai4
+sai5
+san1
+san2
+san3
+san4
+sang1
+sang2
+sang3
+sang4
+sao1
+sao2
+sao3
+sao4
+sao5
+se4
+se5
+sen1
+seng1
+sha1
+sha2
+sha3
+sha4
+sha5
+shai1
+shai4
+shai5
+shan1
+shan2
+shan3
+shan4
+shang1
+shang2
+shang3
+shang4
+shang5
+shao1
+shao2
+shao3
+shao4
+shao5
+she1
+she2
+she3
+she4
+shei2
+shen1
+shen2
+shen3
+shen4
+shen5
+sheng1
+sheng2
+sheng3
+sheng4
+sheng5
+shi1
+shi2
+shi3
+shi4
+shi5
+shir4
+shou1
+shou2
+shou3
+shou4
+shou5
+shu1
+shu2
+shu3
+shu4
+shu5
+shua1
+shua2
+shua3
+shuai1
+shuai2
+shuai3
+shuai4
+shuai5
+shuan1
+shuan4
+shuan5
+shuang1
+shuang2
+shuang3
+shui2
+shui3
+shui4
+shun2
+shun3
+shun4
+shuo1
+shuo4
+shuo5
+si1
+si2
+si3
+si4
+si5
+sil
+song1
+song2
+song3
+song4
+sou1
+sou3
+sou4
+sou5
+sp1
+su1
+su2
+su4
+su5
+suan1
+suan4
+suan5
+sui1
+sui2
+sui3
+sui4
+sun1
+sun2
+sun3
+suo1
+suo2
+suo3
+suo5
+ta1
+ta2
+ta3
+ta4
+ta5
+tai1
+tai2
+tai4
+tai5
+tan1
+tan2
+tan3
+tan4
+tan5
+tang1
+tang2
+tang3
+tang4
+tang5
+tanr1
+tao1
+tao2
+tao3
+tao4
+tao5
+te4
+tei1
+teng2
+teng5
+ter4
+ti1
+ti2
+ti3
+ti4
+ti5
+tian1
+tian2
+tian3
+tianr1
+tiao1
+tiao2
+tiao3
+tiao4
+tie1
+tie2
+tie3
+tie4
+ting1
+ting2
+ting3
+ting5
+tong1
+tong2
+tong3
+tong4
+tong5
+tongr4
+tou1
+tou2
+tou4
+tou5
+tour2
+tu1
+tu2
+tu3
+tu4
+tu5
+tuan1
+tuan2
+tuan3
+tuan4
+tuan5
+tui1
+tui2
+tui3
+tui4
+tui5
+tun1
+tun2
+tun5
+tuo1
+tuo2
+tuo3
+tuo4
+tuo5
+tuor3
+wa1
+wa2
+wa3
+wa4
+wa5
+wai1
+wai2
+wai3
+wai4
+wan1
+wan2
+wan3
+wan4
+wan5
+wang1
+wang2
+wang3
+wang4
+wang5
+wanr1
+wanr2
+wei1
+wei2
+wei3
+wei4
+wei5
+weir4
+wen1
+wen2
+wen3
+wen4
+wen5
+weng1
+weng2
+weng3
+weng4
+wo1
+wo2
+wo3
+wo4
+wo5
+wu1
+wu2
+wu3
+wu4
+wu5
+xi1
+xi2
+xi3
+xi4
+xi5
+xia1
+xia2
+xia4
+xia5
+xian1
+xian2
+xian3
+xian4
+xiang1
+xiang2
+xiang3
+xiang4
+xiao1
+xiao2
+xiao3
+xiao4
+xiao5
+xie1
+xie2
+xie3
+xie4
+xie5
+xin1
+xin4
+xin5
+xing1
+xing2
+xing3
+xing4
+xing5
+xiong1
+xiong2
+xiong5
+xiu1
+xiu3
+xiu4
+xu1
+xu2
+xu3
+xu4
+xu5
+xuan1
+xuan2
+xuan3
+xuan4
+xue1
+xue2
+xue3
+xue4
+xun1
+xun2
+xun4
+ya1
+ya2
+ya3
+ya4
+ya5
+yan1
+yan2
+yan3
+yan4
+yang1
+yang2
+yang3
+yang4
+yang5
+yangr4
+yanr2
+yanr3
+yao1
+yao2
+yao3
+yao4
+yao5
+ye1
+ye2
+ye3
+ye4
+ye5
+yi1
+yi2
+yi3
+yi4
+yi5
+yin1
+yin2
+yin3
+yin4
+ying1
+ying2
+ying3
+ying4
+ying5
+yingr3
+yir4
+yir5
+yo1
+yo5
+yong1
+yong2
+yong3
+yong4
+yong5
+you1
+you2
+you3
+you4
+you5
+yu1
+yu2
+yu3
+yu4
+yu5
+yuan1
+yuan2
+yuan3
+yuan4
+yuan5
+yuanr4
+yue1
+yue4
+yue5
+yun1
+yun2
+yun3
+yun4
+za1
+za2
+za3
+zai1
+zai2
+zai3
+zai4
+zai5
+zan1
+zan2
+zan3
+zan4
+zang1
+zang4
+zao1
+zao2
+zao3
+zao4
+zao5
+zaor3
+ze2
+ze4
+zei2
+zen2
+zen3
+zeng1
+zeng4
+zha1
+zha2
+zha3
+zha4
+zha5
+zhai1
+zhai2
+zhai3
+zhai4
+zhan1
+zhan2
+zhan3
+zhan4
+zhang1
+zhang2
+zhang3
+zhang4
+zhao1
+zhao2
+zhao3
+zhao4
+zhe1
+zhe2
+zhe3
+zhe4
+zhe5
+zhei4
+zhen1
+zhen2
+zhen3
+zhen4
+zheng1
+zheng2
+zheng3
+zheng4
+zheng5
+zher4
+zhi1
+zhi2
+zhi3
+zhi4
+zhi5
+zhong1
+zhong2
+zhong3
+zhong4
+zhou1
+zhou2
+zhou3
+zhou4
+zhou5
+zhu1
+zhu2
+zhu3
+zhu4
+zhu5
+zhua1
+zhua3
+zhuai4
+zhuan1
+zhuan2
+zhuan3
+zhuan4
+zhuan5
+zhuang1
+zhuang4
+zhuang5
+zhui1
+zhui4
+zhun1
+zhun2
+zhun3
+zhunr3
+zhuo1
+zhuo2
+zhur3
+zi1
+zi2
+zi3
+zi4
+zi5
+zong1
+zong2
+zong3
+zong4
+zong5
+zou1
+zou2
+zou3
+zou4
+zu1
+zu2
+zu3
+zu5
+zuan1
+zuan2
+zuan3
+zuan4
+zui2
+zui3
+zui4
+zun1
+zuo1
+zuo2
+zuo3
+zuo4
+zuo5

gp.vocab

@@ -0,0 +1,4131 @@
+啊
+阿
+哀
+哎
+埃
+挨
+捱
+暧
+爱
+癌
+矮
+碍
+艾
+蔼
+隘
+霭
+俺
+唵
+安
+岸
+按
+暗
+案
+氨
+胺
+铵
+鞍
+黯
+昂
+盎
+肮
+傲
+凹
+坳
+奥
+懊
+拗
+敖
+澳
+熬
+獒
+翱
+袄
+八
+叭
+吧
+坝
+峇
+巴
+扒
+把
+拔
+灞
+爸
+疤
+笆
+粑
+罢
+芭
+跋
+霸
+靶
+拜
+掰
+摆
+擘
+柏
+白
+百
+败
+伴
+办
+半
+坂
+扮
+扳
+拌
+搬
+斑
+板
+版
+班
+瓣
+癍
+般
+颁
+傍
+帮
+棒
+榜
+磅
+绑
+膀
+蚌
+谤
+邦
+镑
+保
+包
+堡
+宝
+报
+抱
+暴
+煲
+爆
+胞
+苞
+葆
+薄
+豹
+雹
+饱
+鲍
+倍
+北
+卑
+呗
+备
+悲
+惫
+杯
+狈
+碑
+背
+蓓
+被
+贝
+辈
+奔
+本
+犇
+笨
+苯
+嘣
+崩
+泵
+甭
+绷
+蹦
+迸
+壁
+币
+弊
+彼
+必
+比
+毕
+毙
+璧
+痹
+碧
+笔
+臂
+蔽
+逼
+避
+闭
+鼻
+便
+匾
+变
+扁
+编
+蝙
+贬
+辨
+辩
+边
+遍
+鞭
+俵
+彪
+摽
+标
+表
+裱
+镖
+飙
+鳔
+别
+憋
+瘪
+鳖
+宾
+彬
+斌
+槟
+殡
+滨
+缤
+鬓
+丙
+兵
+冰
+并
+柄
+炳
+病
+秉
+饼
+亳
+伯
+剥
+勃
+博
+卜
+帛
+拨
+搏
+播
+波
+渤
+玻
+礴
+箔
+簸
+脖
+膊
+舶
+菠
+跛
+饽
+驳
+不
+哺
+埠
+布
+怖
+捕
+步
+瓿
+簿
+补
+部
+嚓
+擦
+礤
+偲
+寀
+彩
+才
+材
+猜
+睬
+菜
+蔡
+裁
+财
+踩
+采
+参
+孱
+惨
+惭
+掺
+残
+灿
+璨
+蚕
+餐
+仓
+沧
+舱
+苍
+藏
+嘈
+操
+曹
+槽
+糙
+草
+侧
+册
+厕
+测
+策
+岑
+涔
+噌
+层
+曾
+蹭
+叉
+察
+岔
+差
+插
+杈
+查
+茶
+衩
+诧
+拆
+柴
+豺
+钗
+产
+忏
+搀
+浐
+禅
+缠
+蝉
+谄
+铲
+阐
+颤
+馋
+倡
+偿
+厂
+唱
+场
+娼
+嫦
+尝
+常
+敞
+昌
+猖
+畅
+肠
+吵
+嘲
+巢
+抄
+朝
+潮
+炒
+超
+钞
+彻
+扯
+掣
+撤
+澈
+车
+嗔
+尘
+晨
+沉
+琛
+碜
+臣
+衬
+谌
+趁
+辰
+郴
+陈
+丞
+乘
+呈
+城
+惩
+成
+承
+撑
+晟
+橙
+澄
+瞠
+秤
+称
+程
+诚
+逞
+骋
+侈
+叱
+吃
+尺
+持
+敕
+斥
+池
+炽
+痴
+翅
+耻
+螭
+赤
+迟
+驰
+齿
+充
+冲
+宠
+崇
+虫
+丑
+仇
+愁
+抽
+瞅
+稠
+筹
+绸
+臭
+踌
+酬
+储
+出
+初
+厨
+处
+搐
+楚
+橱
+畜
+础
+触
+躇
+锄
+除
+雏
+啜
+揣
+踹
+串
+传
+喘
+川
+穿
+舛
+船
+创
+幢
+床
+怆
+疮
+窗
+闯
+吹
+垂
+捶
+椎
+槌
+炊
+锤
+唇
+春
+淳
+纯
+蠢
+醇
+戳
+绰
+辍
+龊
+刺
+呲
+慈
+次
+此
+瓷
+疵
+磁
+祠
+糍
+茨
+词
+赐
+辞
+雌
+丛
+从
+匆
+囱
+淙
+聪
+葱
+凑
+促
+殂
+猝
+簇
+粗
+醋
+撺
+窜
+篡
+蹿
+催
+崔
+悴
+摧
+璀
+粹
+翠
+脆
+萃
+存
+寸
+忖
+村
+厝
+嵯
+挫
+措
+搓
+撮
+痤
+矬
+磋
+蹉
+错
+哒
+大
+打
+搭
+沓
+瘩
+答
+耷
+褡
+跶
+达
+靼
+鞑
+代
+傣
+呆
+带
+待
+怠
+戴
+歹
+袋
+贷
+逮
+黛
+丹
+但
+单
+弹
+担
+旦
+殚
+淡
+耽
+胆
+蛋
+诞
+郸
+党
+当
+挡
+档
+荡
+铛
+倒
+刀
+到
+叨
+导
+岛
+悼
+捣
+盗
+祷
+稻
+蹈
+道
+嘚
+得
+德
+的
+凳
+灯
+登
+瞪
+等
+蹬
+邓
+低
+地
+堤
+帝
+底
+弟
+抵
+敌
+滴
+狄
+笛
+第
+缔
+翟
+荻
+蒂
+诋
+迪
+递
+典
+垫
+奠
+巅
+店
+惦
+掂
+殿
+淀
+滇
+点
+电
+甸
+癫
+碘
+颠
+凋
+刁
+叼
+吊
+屌
+掉
+调
+貂
+钓
+雕
+叠
+哋
+嗲
+爹
+牒
+碟
+耋
+蝶
+谍
+跌
+迭
+丁
+叮
+定
+盯
+订
+酊
+钉
+铤
+顶
+鼎
+丢
+东
+冬
+冻
+动
+峒
+懂
+栋
+洞
+董
+兜
+抖
+斗
+痘
+窦
+豆
+逗
+都
+陡
+嘟
+堵
+妒
+度
+杜
+毒
+渡
+犊
+独
+督
+睹
+肚
+读
+赌
+镀
+断
+段
+短
+端
+锻
+兑
+堆
+对
+队
+吨
+囤
+墩
+敦
+沌
+炖
+盹
+盾
+趸
+蹲
+顿
+剁
+哆
+垛
+堕
+多
+夺
+惰
+掇
+朵
+舵
+跺
+踱
+躲
+俄
+厄
+呃
+娥
+婀
+峨
+恶
+愕
+扼
+讹
+遏
+鄂
+颚
+额
+饿
+鳄
+鹅
+诶
+恩
+摁
+二
+儿
+尔
+洱
+而
+耳
+饵
+乏
+伐
+发
+法
+珐
+罚
+阀
+凡
+反
+帆
+梵
+樊
+泛
+烦
+犯
+番
+繁
+翻
+范
+贩
+返
+饭
+仿
+坊
+妨
+房
+放
+方
+纺
+肪
+舫
+芳
+访
+防
+匪
+吠
+啡
+妃
+废
+斐
+狒
+绯
+肥
+肺
+菲
+诽
+费
+霏
+非
+飞
+份
+分
+吩
+坟
+奋
+愤
+氛
+汾
+焚
+粉
+粪
+纷
+芬
+丰
+冯
+凤
+奉
+封
+峰
+枫
+沣
+疯
+缝
+蜂
+讽
+逢
+锋
+风
+否
+付
+伏
+佛
+俘
+俯
+傅
+副
+复
+夫
+妇
+孵
+富
+幅
+府
+扶
+抚
+拂
+敷
+斧
+服
+浮
+涪
+父
+甫
+福
+符
+肤
+腐
+腹
+芙
+蝠
+袱
+覆
+负
+赋
+赴
+辅
+辐
+阜
+附
+伽
+嘎
+噶
+尕
+尬
+旮
+丐
+垓
+改
+概
+溉
+盖
+该
+钙
+尴
+干
+感
+敢
+杆
+柑
+甘
+秆
+竿
+肝
+赣
+赶
+冈
+刚
+岗
+杠
+港
+缸
+钢
+告
+搞
+睾
+稿
+糕
+羔
+膏
+高
+个
+割
+各
+咯
+哥
+嗝
+戈
+搁
+格
+歌
+疙
+胳
+葛
+铬
+镉
+阁
+隔
+革
+骼
+鸽
+给
+亘
+根
+艮
+跟
+哽
+埂
+庚
+更
+梗
+羹
+耕
+耿
+供
+公
+共
+功
+宫
+工
+巩
+弓
+恭
+拱
+攻
+肱
+蚣
+贡
+龚
+勾
+垢
+够
+构
+枸
+沟
+狗
+苟
+购
+钩
+估
+古
+呱
+咕
+固
+姑
+孤
+故
+毂
+股
+菇
+谷
+锢
+雇
+顾
+骨
+鸪
+鼓
+刮
+剐
+卦
+寡
+挂
+瓜
+乖
+怪
+拐
+关
+冠
+官
+惯
+棺
+灌
+管
+罐
+莞
+观
+贯
+馆
+光
+广
+胱
+逛
+傀
+归
+柜
+桂
+瑰
+癸
+硅
+规
+诡
+贵
+跪
+轨
+闺
+鬼
+龟
+棍
+滚
+国
+果
+椁
+裹
+过
+郭
+锅
+哈
+蛤
+亥
+嗨
+孩
+害
+海
+还
+骸
+函
+含
+喊
+寒
+悍
+憨
+憾
+捍
+撼
+旱
+汉
+汗
+涵
+瀚
+焊
+罕
+翰
+蚶
+邯
+酣
+阚
+韩
+鼾
+夯
+杭
+沆
+航
+号
+嚎
+壕
+好
+昊
+毫
+浩
+濠
+皓
+耗
+蚝
+豪
+郝
+何
+合
+呵
+和
+喝
+核
+河
+盒
+禾
+荷
+褐
+贺
+赫
+颌
+鹤
+嘿
+黑
+很
+恨
+狠
+痕
+亨
+哼
+恒
+横
+衡
+吽
+哄
+宏
+弘
+泓
+洪
+烘
+红
+虹
+讧
+轰
+鸿
+侯
+候
+厚
+后
+吼
+喉
+猴
+乎
+互
+呼
+唬
+囫
+壶
+忽
+惚
+户
+护
+沪
+湖
+狐
+瑚
+糊
+胡
+葫
+虎
+蝴
+鹕
+划
+化
+华
+哗
+嬅
+滑
+猾
+画
+花
+话
+坏
+徊
+怀
+槐
+淮
+踝
+唤
+幻
+患
+换
+桓
+欢
+浣
+涣
+焕
+环
+痪
+缓
+凰
+幌
+徨
+恍
+慌
+晃
+湟
+滉
+潢
+煌
+熀
+皇
+磺
+簧
+荒
+蝗
+谎
+黄
+会
+回
+徽
+恢
+悔
+惠
+慧
+挥
+晖
+晦
+毁
+汇
+洄
+灰
+珲
+秽
+绘
+翚
+蕙
+讳
+贿
+辉
+婚
+昏
+浑
+混
+馄
+魂
+伙
+惑
+或
+活
+火
+祸
+获
+豁
+货
+霍
+伎
+冀
+几
+击
+剂
+即
+及
+吉
+唧
+圾
+基
+妓
+姬
+嫉
+季
+寂
+寄
+岌
+己
+忌
+急
+戟
+技
+挤
+既
+暨
+机
+极
+棘
+汲
+济
+激
+犄
+疾
+祭
+积
+箕
+籍
+级
+纪
+继
+绩
+缉
+肌
+脊
+藉
+计
+讥
+记
+辑
+迹
+际
+集
+饥
+骥
+髻
+鲫
+鸡
+价
+佳
+假
+加
+嘉
+夹
+嫁
+家
+戛
+架
+甲
+稼
+茄
+贾
+迦
+钾
+颊
+驾
+件
+俭
+健
+兼
+减
+剑
+剪
+囝
+坚
+奸
+尖
+建
+拣
+捡
+柬
+检
+毽
+涧
+渐
+煎
+监
+简
+箭
+肩
+舰
+艰
+荐
+见
+謇
+贱
+践
+鉴
+锏
+键
+间
+僵
+匠
+奖
+姜
+将
+桨
+江
+浆
+疆
+绛
+缰
+蒋
+讲
+酱
+降
+交
+佼
+侥
+剿
+叫
+娇
+搅
+教
+椒
+浇
+焦
+狡
+皎
+矫
+礁
+绞
+缴
+胶
+脚
+蕉
+蛟
+角
+跤
+轿
+较
+郊
+酵
+铰
+饺
+骄
+介
+借
+劫
+嗟
+姐
+婕
+孑
+届
+戒
+截
+捷
+接
+揭
+杰
+洁
+界
+皆
+睫
+秸
+竭
+结
+节
+街
+解
+诫
+阶
+仅
+今
+劲
+尽
+巾
+斤
+晋
+津
+浸
+瑾
+禁
+筋
+紧
+缙
+襟
+谨
+近
+进
+金
+锦
+靳
+井
+京
+净
+境
+径
+惊
+敬
+景
+晶
+璟
+痉
+睛
+竞
+竟
+精
+经
+茎
+荆
+警
+镜
+靖
+静
+颈
+鲸
+炯
+窘
+迥
+久
+九
+咎
+就
+揪
+救
+旧
+柩
+疚
+究
+纠
+臼
+舅
+酒
+韭
+举
+俱
+具
+剧
+句
+局
+居
+巨
+惧
+拒
+拘
+据
+掬
+桔
+橘
+沮
+炬
+狙
+矩
+聚
+菊
+虡
+趄
+距
+飓
+卷
+娟
+捐
+眷
+倔
+决
+厥
+噘
+嚼
+掘
+撅
+绝
+觉
+诀
+俊
+军
+君
+均
+峻
+浚
+竣
+菌
+郡
+骏
+卡
+咔
+咖
+喀
+凯
+嘅
+开
+忾
+恺
+慨
+楷
+侃
+刊
+勘
+坎
+堪
+槛
+看
+瞰
+砍
+亢
+康
+慷
+扛
+抗
+尻
+烤
+犒
+考
+铐
+靠
+克
+刻
+可
+咳
+嗑
+坷
+壳
+客
+柯
+棵
+渴
+瞌
+磕
+科
+苛
+课
+颗
+啃
+垦
+恳
+肯
+吭
+坑
+铿
+孔
+恐
+控
+空
+口
+叩
+寇
+扣
+抠
+刳
+哭
+库
+枯
+窟
+苦
+裤
+酷
+垮
+夸
+跨
+块
+快
+宽
+款
+况
+匡
+哐
+框
+狂
+眶
+矿
+筐
+亏
+匮
+奎
+愧
+揆
+溃
+煃
+盔
+睽
+窥
+葵
+蝰
+跬
+逵
+馈
+骙
+魁
+困
+坤
+捆
+昆
+焜
+廓
+扩
+括
+阔
+剌
+啦
+喇
+垃
+拉
+旯
+腊
+蜡
+辣
+邋
+来
+睐
+籁
+莱
+赖
+兰
+岚
+懒
+拦
+揽
+栏
+榄
+滥
+澜
+烂
+篮
+缆
+蓝
+览
+啷
+塱
+廊
+朗
+榔
+浪
+狼
+琅
+郎
+佬
+劳
+姥
+捞
+涝
+潦
+烙
+牢
+老
+乐
+了
+肋
+儡
+勒
+嘞
+垒
+泪
+磊
+类
+累
+羸
+蕾
+雷
+冷
+塄
+崚
+愣
+棱
+楞
+丽
+例
+俐
+俪
+利
+力
+励
+历
+厉
+厘
+吏
+呖
+哩
+喱
+娌
+李
+梨
+漓
+狸
+理
+璃
+砾
+礼
+离
+立
+篱
+粒
+罹
+荔
+莉
+藜
+郦
+里
+隶
+鲤
+鹂
+黎
+俩
+帘
+廉
+怜
+恋
+敛
+涟
+炼
+琏
+练
+联
+脸
+莲
+裢
+连
+链
+两
+亮
+凉
+晾
+梁
+粮
+粱
+良
+谅
+踉
+辆
+量
+僚
+嘹
+寥
+廖
+撩
+料
+疗
+缭
+聊
+辽
+冽
+列
+劣
+咧
+烈
+猎
+裂
+趔
+鬣
+临
+凛
+吝
+廪
+拎
+林
+檩
+淋
+琳
+蔺
+赁
+邻
+霖
+鳞
+麟
+令
+伶
+凌
+另
+姈
+岭
+灵
+玲
+祾
+羚
+聆
+菱
+铃
+陵
+零
+领
+龄
+六
+刘
+柳
+榴
+流
+浏
+溜
+琉
+留
+瘤
+硫
+绺
+遛
+咙
+垄
+拢
+窿
+笼
+聋
+胧
+茏
+陇
+隆
+龙
+喽
+娄
+搂
+楼
+漏
+瘘
+篓
+蝼
+镂
+陋
+卢
+卤
+噜
+庐
+录
+掳
+撸
+炉
+碌
+禄
+芦
+赂
+路
+逯
+陆
+露
+颅
+鲁
+鹭
+鹿
+乱
+卵
+孪
+峦
+挛
+栾
+滦
+銮
+仑
+伦
+囵
+抡
+沦
+论
+轮
+啰
+洛
+珞
+络
+罗
+萝
+落
+螺
+裸
+逻
+锣
+骆
+骡
+侣
+吕
+屡
+履
+律
+捋
+旅
+滤
+率
+绿
+虑
+铝
+驴
+掠
+略
+吗
+嘛
+妈
+孖
+玛
+码
+蚂
+蟆
+马
+骂
+麻
+买
+卖
+埋
+脉
+迈
+霾
+麦
+嫚
+慢
+曼
+满
+漫
+瞒
+蔓
+蛮
+馒
+鳗
+忙
+氓
+盲
+芒
+茫
+莽
+蟒
+冇
+冒
+卯
+峁
+帽
+毛
+泖
+猫
+矛
+耄
+茂
+茅
+茆
+貌
+贸
+铆
+锚
+髦
+么
+妹
+媒
+媚
+寐
+昧
+枚
+梅
+每
+没
+煤
+玫
+眉
+美
+莓
+酶
+霉
+魅
+们
+扪
+门
+闷
+勐
+孟
+懵
+朦
+梦
+猛
+盟
+萌
+蒙
+蜢
+锰
+咪
+密
+幂
+弥
+泌
+洣
+眯
+秘
+米
+蜜
+觅
+迷
+靡
+免
+勉
+娩
+棉
+眠
+绵
+缅
+腼
+面
+喵
+妙
+庙
+描
+淼
+渺
+瞄
+秒
+缈
+苗
+藐
+乜
+咩
+灭
+蔑
+悯
+抿
+敏
+民
+泯
+闽
+冥
+名
+命
+明
+茗
+酩
+铭
+鸣
+谬
+墨
+寞
+抹
+摩
+摸
+摹
+末
+模
+沫
+漠
+磨
+膜
+莫
+蘑
+陌
+馍
+魔
+默
+某
+缪
+谋
+亩
+募
+墓
+姆
+幕
+慕
+拇
+木
+母
+牧
+目
+睦
+穆
+嗯
+吶
+呐
+哪
+娜
+拿
+捺
+纳
+肭
+那
+钠
+乃
+奈
+奶
+孻
+耐
+南
+囡
+楠
+湳
+男
+腩
+蝻
+赧
+难
+囊
+囔
+馕
+垴
+婥
+孬
+恼
+挠
+脑
+闹
+呢
+讷
+内
+馁
+嫩
+恁
+能
+你
+倪
+匿
+妮
+尼
+拟
+昵
+泥
+溺
+腻
+逆
+霓
+年
+念
+拈
+捻
+撵
+碾
+蔫
+辗
+鲶
+黏
+娘
+嬢
+酿
+尿
+袅
+鸟
+孽
+捏
+涅
+聂
+蹑
+镊
+您
+凝
+咛
+宁
+拧
+泞
+妞
+扭
+牛
+纽
+农
+哝
+弄
+浓
+努
+奴
+弩
+怒
+暖
+懦
+挪
+糯
+诺
+女
+虐
+哦
+喔
+噢
+偶
+呕
+怄
+欧
+殴
+沤
+鸥
+啪
+帕
+怕
+杷
+爬
+琶
+葩
+趴
+徘
+拍
+排
+派
+湃
+牌
+判
+叛
+攀
+泮
+潘
+畔
+盘
+盼
+乓
+庞
+彷
+旁
+滂
+胖
+螃
+刨
+咆
+抛
+泡
+炮
+狍
+袍
+跑
+佩
+培
+沛
+胚
+裴
+赔
+配
+陪
+喷
+盆
+嘭
+彭
+怦
+抨
+捧
+朋
+棚
+澎
+烹
+砰
+硼
+碰
+篷
+膨
+蓬
+鹏
+僻
+劈
+匹
+啤
+噼
+坯
+媲
+屁
+批
+披
+枇
+毗
+琵
+疲
+痞
+癖
+皮
+纰
+脾
+蜱
+譬
+辟
+偏
+片
+篇
+翩
+骗
+剽
+嫖
+漂
+瓢
+瞟
+票
+缥
+飘
+撇
+瞥
+品
+拼
+聘
+贫
+频
+乒
+凭
+坪
+屏
+平
+瓶
+苹
+萍
+评
+叵
+坡
+婆
+泊
+泼
+珀
+破
+迫
+鄱
+颇
+魄
+剖
+仆
+圃
+埔
+扑
+普
+曝
+朴
+浦
+溥
+濮
+莆
+菩
+葡
+蒲
+谱
+蹼
+铺
+七
+乞
+亓
+企
+其
+凄
+启
+器
+圻
+奇
+契
+妻
+婍
+岂
+岐
+崎
+弃
+戚
+旗
+期
+杞
+栖
+棋
+欺
+歧
+气
+汽
+泣
+淇
+漆
+玘
+琦
+琪
+砌
+祁
+祈
+綮
+绮
+脐
+起
+迄
+骐
+骑
+鳍
+麒
+齐
+恰
+掐
+洽
+乾
+倩
+前
+千
+堑
+嵌
+欠
+歉
+浅
+潜
+牵
+签
+綪
+茜
+虔
+谦
+谴
+迁
+遣
+钱
+钳
+铅
+黔
+呛
+墙
+强
+戗
+抢
+枪
+羌
+腔
+蔷
+跄
+锵
+乔
+侨
+俏
+峭
+巧
+悄
+憔
+撬
+敲
+桥
+瞧
+窍
+翘
+锹
+鞘
+且
+切
+妾
+怯
+惬
+窃
+亲
+侵
+勤
+嗪
+寝
+揿
+擒
+沁
+琴
+禽
+秦
+芹
+钦
+倾
+卿
+庆
+情
+擎
+晴
+氢
+氰
+清
+磬
+蜻
+请
+轻
+青
+顷
+琼
+穷
+穹
+筇
+丘
+囚
+求
+球
+秋
+糗
+裘
+逑
+邱
+酋
+銶
+鳅
+区
+去
+取
+娶
+屈
+岖
+曲
+渠
+瞿
+觑
+趋
+趣
+躯
+驱
+黢
+全
+券
+劝
+圈
+拳
+权
+泉
+犬
+痊
+蜷
+醛
+颧
+却
+瘸
+确
+缺
+雀
+鹊
+群
+裙
+逡
+冉
+染
+然
+燃
+苒
+嚷
+壤
+攘
+瓤
+穰
+让
+娆
+扰
+绕
+饶
+惹
+热
+人
+仁
+仞
+任
+刃
+忍
+纫
+荏
+认
+韧
+饪
+仍
+扔
+日
+冗
+容
+溶
+绒
+茸
+荣
+蓉
+融
+镕
+媃
+揉
+柔
+肉
+乳
+儒
+入
+如
+汝
+濡
+茹
+褥
+辱
+软
+阮
+瑞
+睿
+芮
+蕊
+蕤
+锐
+润
+弱
+若
+仨
+撒
+洒
+萨
+噻
+塞
+腮
+赛
+三
+伞
+散
+馓
+丧
+嗓
+搡
+桑
+嫂
+扫
+瘙
+臊
+骚
+啬
+涩
+瑟
+色
+森
+僧
+傻
+刹
+厦
+啥
+杀
+沙
+煞
+砂
+纱
+莎
+霎
+鲨
+晒
+筛
+删
+善
+山
+扇
+擅
+珊
+缮
+膳
+蟮
+衫
+讪
+赡
+闪
+陕
+鳝
+上
+伤
+商
+尚
+晌
+裳
+赏
+劭
+勺
+哨
+少
+捎
+梢
+潲
+烧
+稍
+绍
+芍
+邵
+韶
+奢
+射
+摄
+歙
+涉
+社
+舌
+舍
+蛇
+设
+麝
+什
+伸
+呻
+婶
+审
+慎
+沈
+深
+渗
+甚
+申
+神
+绅
+肾
+身
+剩
+升
+圣
+声
+牲
+生
+甥
+盛
+省
+笙
+绳
+胜
+世
+事
+仕
+似
+使
+势
+匙
+十
+史
+嗜
+噬
+士
+失
+奭
+始
+实
+室
+寔
+尸
+屎
+市
+师
+式
+恃
+拭
+拾
+施
+时
+是
+柿
+栻
+氏
+湿
+狮
+矢
+石
+示
+舐
+虱
+蚀
+视
+誓
+识
+试
+诗
+辻
+适
+逝
+释
+食
+饰
+驶
+兽
+受
+售
+守
+寿
+手
+授
+收
+瘦
+首
+书
+叔
+墅
+孰
+属
+抒
+数
+暑
+术
+束
+枢
+树
+梳
+殊
+淑
+漱
+熟
+疏
+竖
+纾
+署
+舒
+蔬
+薯
+蜀
+赎
+输
+述
+黍
+鼠
+刷
+耍
+帅
+摔
+甩
+蟀
+衰
+拴
+栓
+涮
+双
+爽
+霜
+水
+睡
+税
+谁
+吮
+瞬
+顺
+朔
+烁
+硕
+说
+丝
+厮
+司
+嗣
+嘶
+四
+寺
+巳
+思
+撕
+斯
+死
+泗
+祀
+私
+肆
+蛳
+饲
+sil
+宋
+嵩
+怂
+悚
+松
+耸
+讼
+送
+叟
+嗽
+搜
+擞
+艘
+薮
+馊
+sp1
+俗
+塑
+夙
+宿
+溯
+稣
+粟
+素
+肃
+苏
+诉
+速
+算
+蒜
+酸
+岁
+濉
+碎
+穗
+绥
+虽
+遂
+邃
+隋
+随
+隧
+髓
+孙
+损
+榫
+荪
+唆
+唢
+嗦
+所
+琐
+索
+缩
+锁
+他
+塌
+塔
+她
+它
+挞
+榻
+踏
+蹋
+遢
+台
+太
+态
+抬
+枱
+汰
+泰
+肽
+胎
+叹
+坍
+坛
+坦
+忐
+探
+摊
+昙
+毯
+滩
+潭
+炭
+痰
+瘫
+碳
+袒
+覃
+谈
+谭
+贪
+倘
+唐
+堂
+塘
+搪
+棠
+汤
+淌
+烫
+糖
+膛
+螳
+趟
+蹚
+躺
+套
+掏
+桃
+涛
+淘
+滔
+萄
+讨
+逃
+陶
+韬
+饕
+忑
+忒
+特
+滕
+疼
+腾
+藤
+誊
+体
+剃
+剔
+啼
+屉
+惕
+提
+替
+梯
+涕
+缇
+踢
+蹄
+题
+鹈
+填
+天
+恬
+添
+甜
+田
+腆
+舔
+挑
+条
+眺
+窕
+笤
+跳
+帖
+贴
+铁
+餮
+亭
+停
+厅
+听
+婷
+庭
+廷
+挺
+艇
+蜓
+霆
+佟
+同
+彤
+捅
+桐
+桶
+痛
+童
+筒
+统
+通
+铜
+偷
+头
+投
+透
+兔
+凸
+吐
+图
+土
+屠
+徒
+涂
+秃
+突
+途
+团
+彖
+湍
+疃
+推
+腿
+蜕
+退
+颓
+吞
+屯
+暾
+臀
+豚
+饨
+唾
+妥
+庹
+托
+拓
+拖
+沱
+砣
+脱
+跎
+陀
+驮
+驼
+鸵
+哇
+娃
+挖
+洼
+瓦
+腽
+蛙
+袜
+外
+崴
+歪
+万
+丸
+婉
+完
+宛
+弯
+挽
+晚
+湾
+玩
+皖
+碗
+腕
+菀
+蜿
+顽
+亡
+妄
+往
+忘
+旺
+望
+枉
+汪
+王
+网
+罔
+为
+伟
+伪
+位
+卫
+危
+味
+唯
+喂
+围
+圩
+委
+威
+娓
+尾
+嵬
+巍
+帷
+微
+惟
+慰
+未
+渭
+炜
+猥
+畏
+纬
+维
+胃
+苇
+萎
+蔚
+薇
+诿
+谓
+违
+韦
+魏
+吻
+文
+汶
+温
+稳
+紊
+纹
+蚊
+问
+闻
+雯
+嗡
+瓮
+翁
+蓊
+卧
+我
+握
+斡
+沃
+涡
+渥
+窝
+蜗
+龌
+乌
+五
+伍
+侮
+务
+勿
+午
+吴
+吾
+呜
+唔
+妩
+屋
+巫
+悟
+捂
+无
+晤
+梧
+武
+毋
+污
+物
+舞
+芜
+蜈
+诬
+误
+雾
+鹉
+习
+兮
+吸
+唏
+喜
+嘻
+夕
+媳
+嬉
+希
+席
+徙
+息
+悉
+惜
+戏
+昔
+晰
+析
+洗
+溪
+熄
+熙
+牺
+犀
+玺
+皙
+禧
+稀
+系
+细
+羲
+膝
+蜥
+蟋
+袭
+西
+锡
+隙
+下
+侠
+吓
+夏
+峡
+暇
+狭
+瑕
+瞎
+虾
+辖
+霞
+仙
+先
+县
+咸
+嫌
+宪
+弦
+掀
+显
+涎
+献
+现
+睍
+纤
+线
+羡
+腺
+舷
+蚬
+衔
+贤
+闲
+限
+险
+陷
+馅
+鲜
+乡
+享
+像
+厢
+向
+响
+巷
+想
+橡
+湘
+相
+祥
+箱
+翔
+襄
+详
+象
+镶
+项
+飨
+香
+哮
+啸
+嚣
+孝
+小
+效
+晓
+校
+消
+淆
+潇
+硝
+笑
+筱
+肖
+萧
+逍
+销
+霄
+些
+亵
+偕
+写
+协
+卸
+屑
+懈
+挟
+携
+斜
+械
+歇
+泄
+泻
+瀣
+燮
+胁
+蟹
+谐
+谢
+邪
+鞋
+信
+心
+新
+昕
+欣
+炘
+芯
+薪
+衅
+辛
+鑫
+馨
+兴
+刑
+型
+姓
+幸
+形
+性
+星
+杏
+猩
+腥
+行
+邢
+醒
+陉
+兄
+凶
+匈
+汹
+熊
+胸
+雄
+休
+修
+咻
+朽
+秀
+绣
+羞
+袖
+锈
+叙
+吁
+嘘
+墟
+婿
+序
+徐
+旭
+栩
+煦
+絮
+绪
+续
+胥
+蓄
+虚
+许
+诩
+需
+须
+顼
+喧
+宣
+悬
+旋
+暄
+渲
+漩
+炫
+玄
+瑄
+癣
+绚
+轩
+选
+削
+学
+穴
+薛
+血
+雪
+靴
+鳕
+勋
+寻
+巡
+循
+旬
+汛
+熏
+窨
+训
+讯
+询
+迅
+逊
+驯
+丫
+亚
+压
+吖
+呀
+哑
+娅
+崖
+押
+桠
+涯
+牙
+芽
+衙
+讶
+轧
+雅
+鸦
+鸭
+严
+俨
+偃
+厌
+咽
+唁
+堰
+妍
+嫣
+嬿
+宴
+岩
+延
+彦
+掩
+晏
+檐
+沿
+淹
+湮
+演
+炎
+烟
+焰
+焱
+燕
+琰
+盐
+眼
+研
+砚
+艳
+蜒
+言
+闫
+雁
+颜
+验
+仰
+佯
+养
+垟
+央
+恙
+扬
+旸
+杨
+样
+殃
+氧
+泱
+洋
+漾
+疡
+痒
+秧
+羊
+阳
+吆
+咬
+垚
+夭
+妖
+姚
+尧
+幺
+摇
+杳
+榣
+滧
+瑶
+窈
+窑
+耀
+肴
+腰
+舀
+药
+要
+谣
+遥
+邀
+钥
+业
+也
+冶
+叶
+嘢
+噎
+夜
+揶
+晔
+曳
+液
+烨
+爷
+耶
+野
+页
+一
+义
+乙
+亦
+亿
+以
+仪
+伊
+依
+倚
+医
+咦
+壹
+夷
+奕
+姨
+宜
+屹
+已
+异
+弈
+弋
+彝
+役
+忆
+怡
+意
+抑
+易
+椅
+毅
+沂
+溢
+疑
+疫
+痍
+癔
+益
+矣
+祎
+移
+绎
+翌
+翼
+胰
+艺
+薏
+蚁
+蜴
+衣
+裔
+议
+译
+谊
+贻
+轶
+迤
+逸
+遗
+颐
+印
+吟
+因
+垠
+姻
+寅
+尹
+引
+殷
+淫
+瘾
+茵
+荫
+鄞
+铟
+银
+阴
+隐
+音
+饮
+婴
+应
+影
+映
+楹
+樱
+溁
+璎
+盈
+硬
+罂
+英
+荧
+莹
+莺
+营
+萦
+蝇
+赢
+迎
+颖
+鹦
+鹰
+哟
+唷
+喲
+佣
+俑
+勇
+咏
+壅
+庸
+恿
+拥
+永
+泳
+涌
+用
+踊
+邕
+雍
+优
+佑
+又
+友
+右
+呦
+尤
+幼
+幽
+忧
+悠
+攸
+有
+柚
+油
+游
+牖
+犹
+由
+莜
+莠
+诱
+邮
+酉
+釉
+黝
+与
+予
+于
+余
+俞
+喻
+域
+妤
+妪
+娱
+宇
+寓
+屿
+峪
+御
+愈
+愉
+愚
+揄
+榆
+欲
+浴
+淤
+渔
+渝
+煜
+狱
+玉
+瑀
+瑜
+瘀
+禹
+禺
+羽
+育
+腴
+臾
+舆
+虞
+裕
+誉
+语
+谀
+谕
+豫
+迂
+逾
+遇
+郁
+钰
+雨
+预
+驭
+鱼
+元
+冤
+原
+员
+园
+圆
+媛
+嫄
+怨
+愿
+援
+渊
+源
+猿
+缘
+羱
+苑
+袁
+辕
+远
+院
+岳
+悦
+曰
+月
+粤
+约
+越
+跃
+钺
+阅
+云
+允
+匀
+孕
+昀
+晕
+殒
+纭
+耘
+蕴
+赟
+运
+酝
+陨
+韵
+匝
+咂
+咋
+杂
+砸
+仔
+再
+在
+宰
+崽
+栽
+灾
+载
+咱
+攒
+昝
+暂
+簪
+糌
+赞
+錾
+脏
+臧
+葬
+赃
+凿
+噪
+早
+枣
+澡
+灶
+燥
+皂
+糟
+蚤
+躁
+造
+遭
+仄
+则
+择
+泽
+责
+贼
+怎
+增
+赠
+乍
+吒
+咤
+喳
+扎
+札
+柞
+栅
+榨
+渣
+炸
+眨
+砟
+蚱
+诈
+闸
+债
+宅
+寨
+摘
+斋
+窄
+占
+展
+崭
+战
+斩
+毡
+沾
+盏
+瞻
+站
+粘
+绽
+蘸
+詹
+丈
+仗
+帐
+幛
+张
+彰
+掌
+杖
+樟
+涨
+漳
+獐
+璋
+章
+胀
+账
+长
+障
+兆
+召
+找
+招
+昭
+照
+爪
+罩
+肇
+赵
+哲
+啫
+折
+浙
+着
+者
+蔗
+辙
+这
+遮
+鹧
+侦
+圳
+帧
+振
+斟
+枕
+珍
+甄
+疹
+真
+祯
+缜
+诊
+贞
+赈
+针
+镇
+阵
+震
+鸩
+争
+峥
+征
+拯
+挣
+政
+整
+正
+症
+睁
+筝
+蒸
+证
+郑
+铮
+之
+侄
+值
+制
+只
+吱
+址
+峙
+帜
+志
+执
+指
+挚
+支
+旨
+智
+枝
+栉
+植
+止
+殖
+汁
+治
+滞
+炙
+痔
+痣
+直
+知
+秩
+稚
+纸
+织
+置
+职
+肢
+脂
+至
+致
+芝
+蜘
+质
+趾
+中
+仲
+众
+冢
+忠
+种
+终
+肿
+衷
+踵
+重
+钟
+周
+咒
+妯
+宙
+州
+帚
+昼
+洲
+皱
+粥
+肘
+胄
+舟
+轴
+骤
+主
+住
+助
+嘱
+拄
+朱
+柱
+株
+注
+洙
+炷
+烛
+煮
+猪
+珠
+瞩
+祝
+竹
+筑
+著
+蛛
+诸
+逐
+驻
+抓
+拽
+专
+撰
+砖
+篆
+赚
+转
+壮
+妆
+庄
+撞
+桩
+状
+装
+坠
+惴
+缀
+赘
+追
+锥
+准
+谆
+卓
+拙
+捉
+桌
+浊
+茁
+酌
+镯
+兹
+咨
+姊
+姿
+子
+字
+梓
+淄
+渍
+滋
+紫
+自
+訾
+资
+龇
+宗
+总
+棕
+纵
+综
+踪
+鬃
+奏
+揍
+走
+邹
+卒
+族
+祖
+租
+组
+诅
+足
+阻
+攥
+纂
+缵
+钻
+嘴
+最
+罪
+醉
+尊
+遵
+佐
+作
+做
+咗
+坐
+左
+座
+昨
+琢
+Ｂ

mtts/__init__.py

@@ -0,0 +1 @@
+from . import datasets, models, text

mtts/datasets/dataset.py

@@ -0,0 +1,207 @@
+import os
+from typing import List, Union
+
+import numpy as np
+import torch
+from torch import Tensor
+from torch.utils.data import DataLoader, Dataset
+
+from mtts.utils.logging import get_logger
+
+logger = get_logger(__file__)
+
+
+def pad_1D(inputs, PAD=0):
+    def pad_data(x, length, PAD):
+        x_padded = np.pad(x, (0, length - x.shape[0]), mode='constant', constant_values=PAD)
+        return x_padded
+
+    max_len = max((len(x) for x in inputs))
+    padded = np.stack([pad_data(x, max_len, PAD) for x in inputs])
+
+    return padded
+
+
+def pad_2D(inputs, maxlen=None):
+    def pad(x, max_len):
+        PAD = 0
+        if np.shape(x)[0] > max_len:
+            raise ValueError("not max_len")
+
+        s = np.shape(x)[1]
+        x_padded = np.pad(x, (0, max_len - np.shape(x)[0]), mode='constant', constant_values=PAD)
+        return x_padded[:, :s]
+
+    if maxlen:
+        output = np.stack([pad(x, maxlen) for x in inputs])
+    else:
+        max_len = max(np.shape(x)[0] for x in inputs)
+        output = np.stack([pad(x, max_len) for x in inputs])
+
+    return output
+
+
+class Tokenizer:
+    def __init__(self, vocab_file):
+        if vocab_file is None:
+            self.vocab = None
+        else:
+            self.vocab = open(vocab_file).read().split('\n')
+            self.v2i = {c: i for i, c in enumerate(self.vocab)}
+
+    def tokenize(self, text: Union[str, List]) -> Tensor:
+        if self.vocab is None:  # direct mapping
+            if isinstance(text, str):
+                tokens = [int(t) for t in text.split()]
+            else:
+                tokens = [int(t) for t in text]
+
+        else:
+            if isinstance(text, str):
+                tokens = [self.v2i[t] for t in text.split()]
+            else:
+                tokens = [self.v2i[t] for t in text]
+        return torch.tensor(tokens)
+
+
+def read_scp(scp_file):
+    with open(scp_file, 'rt') as f:
+        lines = f.read().split('\n')
+    name2value = {line.split()[0]: line.split()[1:] for line in lines if len(line) > 0}
+    return name2value
+
+
+def check_duplicate(keys):
+    key_set0 = set(keys)
+    duplicate = None
+    if len(keys) != len(key_set0):
+        count = {k: 0 for k in key_set0}
+        for k in keys:
+            count[k] += 1
+            if count[k] >= 2:
+                duplicate = k
+                break
+    return duplicate
+    # raise ValueError('duplicated key detected: {duplicate}')
+
+
+def check_keys(*args) -> None:
+    assert len(args) > 0
+    for kv in args:
+        dup = check_duplicate(list(kv.keys()))
+        if dup:
+            raise ValueError('duplicated key detected: {dup}:{kv[dup]}')
+
+    return None
+
+
+class Dataset(Dataset):
+    def __init__(self, config, split='train'):
+        conf = config['dataset'][split]
+        self.name2wav = read_scp(conf['wav_scp'])
+        self.name2mel = read_scp(conf['mel_scp'])
+        self.name2dur = read_scp(conf['dur_scp'])
+
+        self.config = config
+
+        kv_to_check = [self.name2wav, self.name2mel, self.name2dur]
+
+        self.emb_scps = []
+        self.emb_tokenizers = []
+
+        for key in conf.keys():
+            if key.startswith('emb_type'):
+                name2emb = read_scp(conf[key]['scp'])
+                self.emb_scps += [name2emb]
+                emb_tok = Tokenizer(conf[key]['vocab'])
+                self.emb_tokenizers += [emb_tok]
+                logger.info('processed emb {}'.format(conf[key]['_name']))
+
+        kv_to_check += [name2emb]
+        check_keys(*kv_to_check)
+
+        self.names = [name for name in self.name2mel]
+        mel_size = {name: os.path.getsize(self.name2mel[name][0]) for name in self.names}
+
+        self.names = sorted(self.names, key=lambda x: mel_size[x])
+        logger.info(f'Shape of longest mel: {np.load(self.name2mel[self.names[-1]][0]).shape}')
+        logger.info(f'Shape of shortest mel: {np.load(self.name2mel[self.names[0]][0]).shape}')
+
+    def __len__(self):
+        return len(self.name2wav)
+
+    def __getitem__(self, idx):
+        key = self.names[idx]
+        token_tensor = []
+        for scp, tokenizer in zip(self.emb_scps, self.emb_tokenizers):
+            emb_text = scp[key]
+            tokens = tokenizer.tokenize(emb_text)
+            token_tensor.append(torch.unsqueeze(tokens, 0))
+        token_tensor = torch.cat(token_tensor, 0)
+        mel = np.load(self.name2mel[key][0])
+        if mel.shape[0] == self.config['fbank']['n_mels']:
+            mel = torch.tensor(mel.T)
+        else:
+            mel = torch.tensor(mel)
+
+        duration = torch.tensor([int(d) for d in self.name2dur[key]])
+        return token_tensor, duration, mel
+
+
+def pad_1d_tensor(x, n):
+    if x.shape[0] >= n:
+        return x
+    x = torch.cat([x, torch.zeros((n - x.shape[0], ), dtype=x.dtype)], 0)
+    return x
+
+
+def pad_2d_tensor(x, n):
+    if x.shape[1] >= n:
+        return x
+    x = torch.cat([x, torch.zeros((x.shape[0], n - x.shape[1]), dtype=x.dtype)], 1)
+    return x
+
+
+def pad_mel(x, n):
+    if x.shape[0] >= n:
+        return x
+    x = torch.cat([x, torch.zeros((n - x.shape[0], x.shape[1]), dtype=x.dtype)], 0)
+    return x
+
+
+def collate_fn(batch):
+
+    seq_len = []
+    mel_len = []
+    for (token_tensor, duration, mel) in batch:
+        seq_len.append(duration.shape[-1])
+        mel_len.append(mel.shape[0])
+
+    max_seq_len = max(seq_len)
+    max_mel_len = max(mel_len)
+    durations = []
+    token_tensors = []
+    mels = []
+    for token_tensor, duration, mel in batch:
+        duration = pad_1d_tensor(duration, max_seq_len)
+        durations.append(duration.unsqueeze_(0))
+        token_tensor = pad_2d_tensor(token_tensor, max_seq_len)
+        token_tensors.append(token_tensor.unsqueeze_(1))
+        mel = pad_mel(mel, max_mel_len)
+        mels.append(mel.unsqueeze_(0))
+
+    durations = torch.cat(durations, 0)
+    token_tensors = torch.cat(token_tensors, 1)
+    mels = torch.cat(mels, 0)
+
+    return token_tensors, durations, mels, torch.tensor(seq_len), torch.tensor(mel_len)
+
+
+if __name__ == "__main__":
+    import yaml
+    with open('../../examples/aishell3/config.yaml') as f:
+        config = yaml.safe_load(f)
+    dataset = Dataset(config)
+    dataloader = DataLoader(dataset, batch_size=6, collate_fn=collate_fn)
+    batch = next(iter(dataloader))
+    print(type(batch[-1]))

mtts/loss.py

@@ -0,0 +1,23 @@
+import torch
+import torch.nn as nn
+
+
+class FS2Loss(nn.Module):
+    def __init__(self):
+        super(FS2Loss, self).__init__()
+        self.mse_loss = nn.MSELoss()
+        self.mae_loss = nn.L1Loss()
+
+    def forward(self, d_pred, d_truth, mel_pred, mel_postnet, mel_truth, src_mask, mel_mask):
+        d_pred = d_pred.masked_select(src_mask)
+        d_truth = d_truth.masked_select(src_mask)
+
+        mel_pred = mel_pred.masked_select(mel_mask.unsqueeze(-1))
+        mel_postnet = mel_postnet.masked_select(mel_mask.unsqueeze(-1))
+        mel_truth = mel_truth.masked_select(mel_mask.unsqueeze(-1))
+
+        mel_loss = self.mse_loss(mel_pred, mel_truth) * 0.1
+        mel_postnet_loss = self.mse_loss(mel_postnet, mel_truth)
+        d_loss = self.mae_loss(d_pred, d_truth) * 0.01
+
+        return mel_loss, mel_postnet_loss, d_loss

mtts/models/__init__.py

@@ -0,0 +1 @@
+from . import decoder, encoder, postnet, vocoder

mtts/models/decoder/__init__.py

@@ -0,0 +1,2 @@
+from .decoder import Decoder
+from .fs2_transformer_decoder import FS2TransformerDecoder

mtts/models/decoder/decoder.py

@@ -0,0 +1,22 @@
+import torch.nn as nn
+
+from mtts.models.decoder.fs2_transformer_decoder import FS2TransformerDecoder
+from mtts.utils.logging import get_logger
+
+logger = get_logger(__file__)
+DECODERS = [FS2TransformerDecoder]
+
+
+class Decoder(nn.Module):
+    """
+    """
+    def __init__(self, decoder_type: str = 'FS2TransformerDecoder', **kwargs):
+        super(Decoder, self).__init__()
+        logger.info(f'decoder_type {decoder_type}')
+        decoder_class = eval(decoder_type)
+        assert decoder_class in DECODERS
+
+        self.decoder = decoder_class(**kwargs)
+
+    def forward(self, *args, **kwargs):
+        return self.decoder(*args, **kwargs)

mtts/models/decoder/fs2_transformer_decoder.py

@@ -0,0 +1,67 @@
+import torch.nn as nn
+from torch import Tensor
+
+from ..layers import FFTBlock, get_sinusoid_encoding_table
+
+
+class FS2TransformerDecoder(nn.Module):
+    """ A decoder that accepts a list of sequences as input and
+    out a sequence as output.  The input and output sequences share the same length
+
+    The input sequence is a list of tensors, which may contain text-embedding, speaker-embeddings.
+
+
+    """
+    def __init__(
+            self,
+            input_dim: int = 256,  # must ==  decoder output dim
+            n_layers: int = 4,
+            n_heads: int = 2,
+            hidden_dim: int = 256,
+            d_inner: int = 1024,
+            dropout: float = 0.5,
+            max_len: int = 2048,  # for computing position table
+    ):
+        super(FS2TransformerDecoder, self).__init__()
+
+        self.input_dim = input_dim
+        self.input_dim = input_dim
+
+        self.max_len = max_len
+
+        d_k = hidden_dim // n_heads
+        d_v = hidden_dim // n_heads
+
+        n_position = max_len + 1
+
+        # self.speaker_fc = nn.Linear(512, 256, bias=False)
+
+        pos_table = get_sinusoid_encoding_table(n_position, input_dim).unsqueeze(0)
+
+        self.position_enc = nn.Parameter(pos_table, requires_grad=False)
+        layers = []
+        for _ in range(n_layers):
+            layer = FFTBlock(hidden_dim, d_inner, n_heads, d_k, d_v, dropout=dropout)
+            layers.append(layer)
+
+        self.layers = nn.ModuleList(layers)
+
+    def forward(self, input: Tensor, mask):
+        batch_size, seq_len, input_dim = input.shape[0:3]
+        if input.shape[1] != seq_len:
+            raise ValueError('The input sequences must have the same length')
+        if input.shape[1] != seq_len:
+            raise ValueError('The input sequences must have the same dimension')
+
+        attn_mask = mask.unsqueeze(1).expand(-1, seq_len, -1)
+
+        if input.shape[1] > self.max_len:
+            raise ValueError('inputs.shape[1] > self.max_len')
+
+        pos_embed = self.position_enc[:, :seq_len, :].expand(batch_size, -1, -1)
+
+        output = input + pos_embed
+        for layer in self.layers:
+            output, dec_slf_attn = layer(output, mask=mask, attn_mask=attn_mask)
+
+        return output

mtts/models/encoder/__init__.py

@@ -0,0 +1,2 @@
+from .encoder import Encoder
+from .fs2_transformer_encoder import FS2TransformerEncoder

mtts/models/encoder/encoder.py

@@ -0,0 +1,25 @@
+# from mtts.models.fs2_model import ENCODERS
+import torch.nn as nn
+
+from mtts.utils.logging import get_logger
+
+from .fs2_transformer_encoder import FS2TransformerEncoder
+
+logger = get_logger(__file__)
+
+ENCODERS = [FS2TransformerEncoder]
+
+
+class Encoder(nn.Module):
+    ''' Encoder '''
+    def __init__(self, encoder_type: str = 'FS2TransformerEncoder', **kwargs):
+
+        super(Encoder, self).__init__()
+        logger.info(f'building encoder with type:{encoder_type}')
+        encoder_class = eval(encoder_type)
+        assert encoder_class in ENCODERS
+        self.config = kwargs
+        self.encoder = encoder_class(**kwargs)
+
+    def forward(self, *args, **kwargs):
+        return self.encoder(*args, **kwargs)

mtts/models/encoder/fs2_transformer_encoder.py

@@ -0,0 +1,65 @@
+from typing import List
+
+import torch.nn as nn
+from torch import Tensor
+
+from ..layers import FFTBlock, get_sinusoid_encoding_table
+
+
+class FS2TransformerEncoder(nn.Module):
+    ''' FS2TransformerEncoder '''
+    def __init__(
+        self,
+        emb_layers: nn.ModuleList,
+        embeding_weights: List[float],
+        hidden_dim: int = 256,
+        n_layers: int = 4,
+        n_heads: int = 2,
+        d_inner: int = 1024,
+        dropout: float = 0.5,
+        max_len: int = 1024,
+    ):
+
+        super(FS2TransformerEncoder, self).__init__()
+
+        self.emb_layers = emb_layers
+        self.embeding_weights = embeding_weights
+        self.hidden_dim = hidden_dim
+
+        d_k = hidden_dim // n_heads
+        d_v = hidden_dim // n_heads
+        d_model = hidden_dim
+        self.layers = nn.ModuleList()
+        for _ in range(n_layers):
+            layer = FFTBlock(d_model, d_inner, n_heads, d_k, d_v, dropout=dropout)
+            self.layers.append(layer)
+
+    def forward(self, texts: List[Tensor], mask: Tensor):
+
+        if len(self.embeding_weights) != len(texts):
+            raise ValueError(f'Input texts has length {len(texts)},　\
+                    but embedding module list has length {len(self.embeding_weights)}')
+        batch_size = texts[0].shape[0]
+        seq_len = texts[0].shape[1]
+        attn_mask = mask.unsqueeze(1).expand(-1, seq_len, -1)
+        text_embed = self.emb_layers[0](texts[0]) * self.embeding_weights[0]
+        n_embs = len(self.embeding_weights)
+        for i in range(1, n_embs):
+            text_embed += self.emb_layers[i](texts[i]) * self.embeding_weights[i]
+
+        if self.training:
+            pos_embed = get_sinusoid_encoding_table(seq_len, self.hidden_dim)
+            assert pos_embed.shape[0] == seq_len
+            pos_embed = pos_embed[:seq_len, :]
+            pos_embed = pos_embed.unsqueeze(0).expand(batch_size, -1, -1)
+            pos_embed = pos_embed.to(texts[0].device)
+        else:
+            pos_embed = self.position_enc[:, :seq_len, :]
+            pos_embed = pos_embed.expand(batch_size, -1, -1)
+
+        all_embed = text_embed + pos_embed
+
+        for layer in self.layers:
+            all_embed, enc_slf_attn = layer(all_embed, mask=mask, attn_mask=attn_mask)
+
+        return all_embed

mtts/models/fs2_model.py

@@ -0,0 +1,135 @@
+from typing import List, Optional
+
+import torch
+import torch.nn as nn
+from torch import Tensor
+
+from mtts.models.decoder import Decoder
+from mtts.models.encoder import FS2TransformerEncoder
+from mtts.models.fs2_variance import VarianceAdaptor
+from mtts.models.postnet import PostNet
+from mtts.utils.logging import get_logger
+
+ENCODERS = [
+    FS2TransformerEncoder,
+]
+
+logger = get_logger(__file__)
+
+
+def __read_vocab(file):
+    with open(file) as f:
+        lines = f.read().split('\n')
+    lines = [line for line in lines if len(line) > 0]
+    return lines
+
+
+def _get_layer(emb_config: dict):  # -> Optional[List[nn.Module],List[float]]:
+    logger.info(f'building embedding with config: {emb_config}')
+    if emb_config['enable']:
+        if emb_config['vocab'] is None:
+            vocab_size = emb_config['vocab_size']
+        else:
+            vocab = __read_vocab(emb_config['vocab'])
+            vocab_size = len(vocab)
+        layer = nn.Embedding(vocab_size, emb_config['dim'], padding_idx=0)
+        return layer, emb_config['weight']
+    else:
+        return None, None
+
+
+def _build_embedding_layers(config):
+    layers = nn.ModuleList()
+    weights = []
+    for c in (
+            config['pinyin_embedding'],
+            config['hanzi_embedding'],
+            config['speaker_embedding'],
+    ):
+
+        layer, weight = _get_layer(c)
+        if layer is not None:
+            layers.append(layer)
+            weights.append(weight)
+
+    return layers, weights
+
+
+def get_mask_from_lengths(lengths, max_len=None):
+    batch_size = lengths.shape[0]
+    if max_len is None:
+        max_len = torch.max(lengths).item()
+    ids = torch.arange(0, max_len).unsqueeze(0).expand(batch_size, -1).to(lengths.device)
+    mask = (ids >= lengths.unsqueeze(1).expand(-1, max_len))
+
+    return mask
+
+
+class FastSpeech2(nn.Module):
+    """ FastSpeech2 """
+    def __init__(self, config):
+        super(FastSpeech2, self).__init__()
+
+        emb_layers, emb_weights = _build_embedding_layers(config)
+
+        EncoderClass = eval(config['encoder']['encoder_type'])
+        assert EncoderClass in ENCODERS
+        encoder_conf = config['encoder']['conf']
+        encoder_conf.update({'emb_layers': emb_layers})
+        encoder_conf.update({'embeding_weights': emb_weights})
+        self.encoder = EncoderClass(**encoder_conf)
+
+        dur_config = config['duration_predictor']
+        self.variance_adaptor = VarianceAdaptor(**dur_config)
+
+        decoder_config = config['decoder']
+        self.decoder = Decoder(**decoder_config)
+
+        n_mels = config['fbank']['n_mels']
+        mel_linear_input_dim = decoder_config['hidden_dim']
+        self.mel_linear = nn.Linear(mel_linear_input_dim, n_mels)
+        self.postnet = PostNet(**config['postnet'], )
+
+    def forward(self,
+                input_seqs: List[Tensor],
+                seq_len: Tensor,
+                mel_len: Optional[Tensor] = None,
+                d_target: Optional[Tensor] = None,
+                max_src_len=None,
+                max_mel_len=None,
+                d_control=1.0,
+                p_control=1.0,
+                e_control=1.0):
+        src_mask = get_mask_from_lengths(seq_len, max_src_len)
+
+        if mel_len is not None:
+            mel_mask = get_mask_from_lengths(mel_len, max_mel_len)
+        else:
+            mel_mask = None
+
+        encoder_output = self.encoder(input_seqs, src_mask)
+
+        if d_target is not None:
+            variance_adaptor_output, d_prediction, _, _ = self.variance_adaptor(encoder_output, src_mask, mel_mask,
+                                                                                d_target, max_mel_len, d_control)
+        else:
+            variance_adaptor_output, d_prediction, mel_len, mel_mask = self.variance_adaptor(
+                encoder_output, src_mask, mel_mask, d_target, max_mel_len, d_control)
+
+        decoder_output = self.decoder(variance_adaptor_output, mel_mask)
+
+        mel_pred = self.mel_linear(decoder_output)
+        postnet_input = torch.unsqueeze(mel_pred, 1)
+        postnet_output = self.postnet(postnet_input) + mel_pred
+
+        return mel_pred, postnet_output, d_prediction, src_mask, mel_mask, mel_len
+
+
+if __name__ == "__main__":
+    # Test
+    import yaml
+    with open('../../examples/aishell3/config.yaml') as f:
+        config = yaml.safe_load(f)
+    model = FastSpeech2(config)
+    print(model)
+    print(sum(param.numel() for param in model.parameters()))

mtts/models/fs2_variance.py

@@ -0,0 +1,189 @@
+from collections import OrderedDict
+from typing import Optional
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from torch import Tensor
+
+
+def get_mask_from_lengths(lengths, max_len=None):
+    batch_size = lengths.shape[0]
+    if max_len is None:
+        max_len = torch.max(lengths).item()
+    ids = torch.arange(0, max_len).unsqueeze(0).expand(batch_size, -1).to(lengths.device)
+    mask = (ids >= lengths.unsqueeze(1).expand(-1, max_len))
+
+    return mask
+
+
+def pad(input_ele, mel_max_length=None):
+    if mel_max_length:
+        max_len = mel_max_length
+    else:
+        max_len = max([input_ele[i].size(0) for i in range(len(input_ele))])
+
+    out_list = list()
+    for i, batch in enumerate(input_ele):
+        if len(batch.shape) == 1:
+            one_batch_padded = F.pad(batch, (0, max_len - batch.size(0)), "constant", 0.0)
+        elif len(batch.shape) == 2:
+            one_batch_padded = F.pad(batch, (0, 0, 0, max_len - batch.size(0)), "constant", 0.0)
+        out_list.append(one_batch_padded)
+    out_padded = torch.stack(out_list)
+    return out_padded
+
+
+# def clones(module, N):
+#     return nn.ModuleList([copy.deepcopy(module) for _ in range(N)])
+
+
+class Conv(nn.Module):
+    """
+    Convolution Module
+    """
+    def __init__(self,
+                 in_channels: int,
+                 out_channels: int,
+                 kernel_size: int = 1,
+                 stride: int = 1,
+                 padding: int = 0,
+                 dilation: int = 1,
+                 bias: bool = True,
+                 w_init: str = 'linear'):
+        """
+        :param in_channels: dimension of input
+        :param out_channels: dimension of output
+        :param kernel_size: size of kernel
+        :param stride: size of stride
+        :param padding: size of padding
+        :param dilation: dilation rate
+        :param bias: boolean. if True, bias is included.
+        :param w_init: str. weight inits with xavier initialization.
+        """
+        super(Conv, self).__init__()
+
+        self.conv = nn.Conv1d(in_channels,
+                              out_channels,
+                              kernel_size=kernel_size,
+                              stride=stride,
+                              padding=padding,
+                              dilation=dilation,
+                              bias=bias)
+
+    def forward(self, x):
+        x = x.contiguous().transpose(1, 2)
+        x = self.conv(x)
+        x = x.contiguous().transpose(1, 2)
+
+        return x
+
+
+class VarianceAdaptor(nn.Module):
+    """ Variance Adaptor """
+    def __init__(self,
+                 duration_mean: float,
+                 input_dim: int = 256,
+                 filter_size: int = 256,
+                 kernel_size: int = 3,
+                 dropout: float = 0.5):
+        super(VarianceAdaptor, self).__init__()
+        self.duration_predictor = VariancePredictor(input_dim, filter_size, kernel_size, dropout)
+        self.length_regulator = LengthRegulator()
+        self.duration_mean = duration_mean
+
+    def forward(self,
+                x: Tensor,
+                src_mask: Tensor,
+                mel_mask: Optional[Tensor] = None,
+                duration_target: Optional[Tensor] = None,
+                max_len: Optional[int] = None,
+                d_control: float = 1.0):
+
+        log_duration_prediction = self.duration_predictor(x, src_mask)
+        if duration_target is not None:
+            duration_rounded = torch.clamp(torch.round((duration_target + self.duration_mean) * d_control), min=0)
+
+            x, mel_len = self.length_regulator(x, duration_rounded, max_len)
+        else:
+            # duration_rounded = torch.clamp(
+            #   (torch.round(torch.exp(log_duration_prediction)-hp.log_offset)*d_control), min=0)
+            duration_rounded = torch.clamp(torch.round(
+                (log_duration_prediction.detach() + self.duration_mean) * d_control),
+                                           min=0)
+            # print('duration',duration_rounded)
+
+            x, mel_len = self.length_regulator(x, duration_rounded, max_len)
+            mel_mask = get_mask_from_lengths(mel_len)
+
+        return x, log_duration_prediction, mel_len, mel_mask
+
+
+class LengthRegulator(nn.Module):
+    """ Length Regulator """
+    def __init__(self):
+        super(LengthRegulator, self).__init__()
+
+    def LR(self, x, duration, max_len):
+        output = list()
+        mel_len = list()
+        for batch, expand_target in zip(x, duration):
+            expanded = self.expand(batch, expand_target)
+            output.append(expanded)
+            mel_len.append(expanded.shape[0])
+
+        if max_len is not None:
+            output = pad(output, max_len)
+        else:
+            output = pad(output)
+
+        return output, torch.LongTensor(mel_len).to(x.device)
+
+    def expand(self, batch, predicted):
+        out = list()
+
+        for i, vec in enumerate(batch):
+            expand_size = predicted[i].item()
+            out.append(vec.expand(int(expand_size), -1))
+        out = torch.cat(out, 0)
+
+        return out
+
+    def forward(self, x, duration, max_len):
+        output, mel_len = self.LR(x, duration, max_len)
+        return output, mel_len
+
+
+class VariancePredictor(nn.Module):
+    """ Duration, Pitch and Energy Predictor """
+    def __init__(self, encoder_dim: int = 256, filter_size: int = 256, kernel_size: int = 3, dropout: float = 0.5):
+        super(VariancePredictor, self).__init__()
+
+        self.input_size = encoder_dim
+        self.filter_size = filter_size
+        self.kernel = kernel_size
+        self.conv_output_size = filter_size
+        self.dropout = dropout
+
+        self.conv_layer = nn.Sequential(
+            OrderedDict([("conv1d_1",
+                          Conv(self.input_size,
+                               self.filter_size,
+                               kernel_size=self.kernel,
+                               padding=(self.kernel - 1) // 2)), ("relu_1", nn.LeakyReLU()),
+                         ("layer_norm_1", nn.LayerNorm(self.filter_size)), ("dropout_1", nn.Dropout(self.dropout)),
+                         ("conv1d_2", Conv(self.filter_size, self.filter_size, kernel_size=self.kernel, padding=1)),
+                         ("relu_2", nn.LeakyReLU()), ("layer_norm_2", nn.LayerNorm(self.filter_size)),
+                         ("dropout_2", nn.Dropout(self.dropout))]))
+
+        self.linear_layer = nn.Linear(self.conv_output_size, 1)
+
+    def forward(self, encoder_output, mask):
+        out = self.conv_layer(encoder_output)
+        out = self.linear_layer(out)
+        out = out.squeeze(-1)
+
+        if mask is not None:
+            out = out.masked_fill(mask, 0.)
+
+        return out

mtts/models/layers.py

@@ -0,0 +1,170 @@
+from typing import Optional
+
+import numpy as np
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from torch import Tensor
+
+
+def get_sinusoid_encoding_table(n_position, d_hid, padding_idx=None):
+    ''' Sinusoid position encoding table '''
+    def cal_angle(position, hid_idx):
+        return position / np.power(10000, 2 * (hid_idx // 2) / d_hid)
+
+    def get_posi_angle_vec(position):
+        return [cal_angle(position, hid_j) for hid_j in range(d_hid)]
+
+    sinusoid_table = np.array([get_posi_angle_vec(pos_i) for pos_i in range(n_position)])
+
+    sinusoid_table[:, 0::2] = np.sin(sinusoid_table[:, 0::2])  # dim 2i
+    sinusoid_table[:, 1::2] = np.cos(sinusoid_table[:, 1::2])  # dim 2i+1
+
+    if padding_idx is not None:
+        # zero vector for padding dimension
+        sinusoid_table[padding_idx] = 0.
+
+    return torch.FloatTensor(sinusoid_table)
+
+
+class ScaledDotProductAttention(nn.Module):
+    ''' Scaled Dot-Product Attention '''
+    def __init__(self, temperature):
+        super().__init__()
+        self.temperature = temperature
+        self.softmax = nn.Softmax(dim=2)
+
+    def forward(self, q, k, v, mask=None):
+
+        attn = torch.bmm(q, k.transpose(1, 2))
+        attn = attn / self.temperature
+        if mask is not None:
+            attn = attn.masked_fill(mask, -np.inf)
+
+        attn = self.softmax(attn)
+        output = torch.bmm(attn, v)
+
+        return output, attn
+
+
+class MultiHeadAttention(nn.Module):
+    ''' Multi-Head Attention module '''
+    def __init__(self, n_head, d_model, d_k, d_v, dropout=0.1):
+        super().__init__()
+
+        self.n_head = n_head
+        self.d_k = d_k
+        self.d_v = d_v
+
+        self.w_qs = nn.Linear(d_model, n_head * d_k)
+        self.w_ks = nn.Linear(d_model, n_head * d_k)
+        self.w_vs = nn.Linear(d_model, n_head * d_v)
+
+        self.attention = ScaledDotProductAttention(temperature=np.power(d_k, 0.5))
+        self.layer_norm = nn.LayerNorm(d_model)
+
+        self.fc = nn.Linear(n_head * d_v, d_model)
+
+        self.dropout = nn.Dropout(dropout)
+
+    def forward(self, q: Tensor, k: Tensor, v: Tensor, mask: Optional[Tensor] = None):
+
+        d_k, d_v, n_head = self.d_k, self.d_v, self.n_head
+
+        sz_b, len_q, _ = q.size()
+        sz_b, len_k, _ = k.size()
+        sz_b, len_v, _ = v.size()
+
+        residual = q
+
+        q = self.w_qs(q).view(sz_b, len_q, n_head, d_k)
+        k = self.w_ks(k).view(sz_b, len_k, n_head, d_k)
+        v = self.w_vs(v).view(sz_b, len_v, n_head, d_v)
+        q = q.permute(2, 0, 1, 3).contiguous().view(-1, len_q, d_k)  # (n*b) x lq x dk
+        k = k.permute(2, 0, 1, 3).contiguous().view(-1, len_k, d_k)  # (n*b) x lk x dk
+        v = v.permute(2, 0, 1, 3).contiguous().view(-1, len_v, d_v)  # (n*b) x lv x dv
+
+        mask = mask.repeat(n_head, 1, 1).to(q.device)
+        output, attn = self.attention(q, k, v, mask=mask)
+
+        output = output.view(n_head, sz_b, len_q, d_v)
+        output = output.permute(1, 2, 0, 3).contiguous().view(sz_b, len_q, -1)  # b x lq x (n*dv)
+
+        output = self.dropout(self.fc(output))
+        output = self.layer_norm(output + residual)
+
+        return output, attn
+
+
+class PositionwiseFeedForward(nn.Module):
+    ''' A two-feed-forward-layer module '''
+    def __init__(self, d_in: int, d_hid: int, kernel_size: int = 9, dropout: float = 0.1):
+        super().__init__()
+
+        # Use Conv1D
+        # position-wise
+        self.w_1 = nn.Conv1d(d_in, d_hid, kernel_size=kernel_size, padding=(kernel_size - 1) // 2)
+        # position-wise
+        self.w_2 = nn.Conv1d(d_hid, d_in, kernel_size=1, padding=0)
+
+        self.layer_norm = nn.LayerNorm(d_in)
+        self.dropout = nn.Dropout(dropout)
+
+    def forward(self, x: Tensor):
+        residual = x
+        output = x.transpose(1, 2)
+        output = self.w_2(F.relu(self.w_1(output)))
+        output = output.transpose(1, 2)
+        output = self.dropout(output)
+        output = self.layer_norm(output + residual)
+
+        return output
+
+
+class FFTBlock(torch.nn.Module):
+    """FFT Block"""
+    def __init__(self, d_model: int, d_inner: int, n_head: int, d_k: int, d_v: int, dropout: float = 0.1):
+        super(FFTBlock, self).__init__()
+
+        self.slf_attn = MultiHeadAttention(n_head, d_model, d_k, d_v, dropout=dropout)
+        self.pos_ffn = PositionwiseFeedForward(d_model, d_inner, dropout=dropout)
+
+    def forward(self, enc_input: Tensor, mask: Optional[Tensor] = None, attn_mask: Optional[Tensor] = None):
+
+        enc_output, enc_slf_attn = self.slf_attn(enc_input, enc_input, enc_input, mask=attn_mask)
+        enc_output = enc_output.masked_fill(mask.unsqueeze(-1), 0)
+
+        enc_output = self.pos_ffn(enc_output)
+        enc_output = enc_output.masked_fill(mask.unsqueeze(-1), 0)
+
+        return enc_output, enc_slf_attn
+
+
+class ConvNorm(torch.nn.Module):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 kernel_size=1,
+                 stride=1,
+                 padding=None,
+                 dilation=1,
+                 bias=True,
+                 w_init_gain='linear'):
+        super(ConvNorm, self).__init__()
+
+        if padding is None:
+            assert (kernel_size % 2 == 1)
+            padding = int(dilation * (kernel_size - 1) / 2)
+
+        self.conv = torch.nn.Conv1d(in_channels,
+                                    out_channels,
+                                    kernel_size=kernel_size,
+                                    stride=stride,
+                                    padding=padding,
+                                    dilation=dilation,
+                                    bias=bias)
+
+    def forward(self, signal):
+        conv_signal = self.conv(signal)
+
+        return conv_signal

mtts/models/postnet/__init__.py

@@ -0,0 +1,3 @@
+from .postnet import PostNet
+from .postnet_1d import PostNet1d
+from .postnet_unet import PostUNet

mtts/models/postnet/postnet.py

@@ -0,0 +1,22 @@
+import torch.nn as nn
+
+from .postnet_1d import PostNet1d
+from .postnet_unet import PostUNet
+
+POSTNETS = [PostNet1d, PostUNet]
+
+
+class PostNet(nn.Module):
+    """
+    Interface class for postnets
+    """
+    def __init__(self, postnet_type: str = 'PostUNet', **kwargs):
+
+        super(PostNet, self).__init__()
+        PostNetClass = eval(postnet_type)
+        assert PostNetClass in POSTNETS
+        self.postnet = PostNetClass(**kwargs)
+        self.config = kwargs
+
+    def forward(self, x, **kwargs):
+        return self.postnet(x, **kwargs)

mtts/models/postnet/postnet_1d.py

@@ -0,0 +1,56 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from ..layers import ConvNorm
+
+
+class PostNet1d(nn.Module):
+    """
+    PostNet: Five 1-d convolution with 512 channels and kernel size 5
+    """
+    def __init__(self, n_mel_channels=80, postnet_embedding_dim=512, postnet_kernel_size=7, postnet_n_convolutions=7):
+
+        super(PostNet1d, self).__init__()
+        self.convolutions = nn.ModuleList()
+
+        self.convolutions.append(
+            nn.Sequential(
+                ConvNorm(n_mel_channels,
+                         postnet_embedding_dim,
+                         kernel_size=postnet_kernel_size,
+                         stride=1,
+                         padding=int((postnet_kernel_size - 1) / 2),
+                         dilation=1,
+                         w_init_gain='tanh'), nn.InstanceNorm1d(postnet_embedding_dim)))
+
+        for i in range(1, postnet_n_convolutions - 1):
+            self.convolutions.append(
+                nn.Sequential(
+                    ConvNorm(postnet_embedding_dim,
+                             postnet_embedding_dim,
+                             kernel_size=postnet_kernel_size,
+                             stride=1,
+                             padding=int((postnet_kernel_size - 1) / 2),
+                             dilation=1,
+                             w_init_gain='tanh'), nn.InstanceNorm1d(postnet_embedding_dim)))
+
+        self.convolutions.append(
+            nn.Sequential(
+                ConvNorm(postnet_embedding_dim,
+                         n_mel_channels,
+                         kernel_size=postnet_kernel_size,
+                         stride=1,
+                         padding=int((postnet_kernel_size - 1) / 2),
+                         dilation=1,
+                         w_init_gain='linear'), nn.InstanceNorm1d(n_mel_channels)))
+
+    def forward(self, x):
+        x = x.contiguous().transpose(1, 2)
+
+        for i in range(len(self.convolutions) - 1):
+            x = F.dropout(torch.tanh(self.convolutions[i](x)), 0.1, self.training)
+        x = F.dropout(self.convolutions[-1](x), 0.1, self.training)
+
+        x = x.contiguous().transpose(1, 2)
+        return x

mtts/models/postnet/postnet_unet.py

@@ -0,0 +1,117 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+class double_res_conv(nn.Module):
+    def __init__(self, in_ch, out_ch, bn=False):
+        super(double_res_conv, self).__init__()
+        self.conv1 = nn.Sequential(
+            nn.Conv2d(in_ch, out_ch, 3, padding=1),
+            nn.InstanceNorm2d(out_ch),
+        )
+        self.conv2 = nn.Sequential(
+            nn.Conv2d(out_ch, out_ch, 3, padding=1),
+            nn.InstanceNorm2d(out_ch),
+        )
+
+        self.relu = nn.LeakyReLU(0.1)
+
+    def forward(self, x):
+
+        x1 = self.conv1(x)
+        x2 = self.conv2(x1)
+        x3 = self.relu(x2)
+
+        return x3
+
+
+class inconv(nn.Module):
+    def __init__(self, in_ch, out_ch, bn=True):
+        super(inconv, self).__init__()
+
+        self.conv = double_res_conv(in_ch, out_ch, bn)
+
+    def forward(self, x):
+
+        x = self.conv(x)
+        return x
+
+
+class down(nn.Module):
+    def __init__(self, in_ch, out_ch, bn=True):
+        super(down, self).__init__()
+        self.mpconv = nn.Sequential(nn.AvgPool2d(2), double_res_conv(in_ch, out_ch, bn))
+
+    def forward(self, x):
+        x = self.mpconv(x)
+        return x
+
+
+class up(nn.Module):
+    def __init__(self, in_ch, out_ch, bilinear=True, bn=True):
+        super(up, self).__init__()
+
+        self.bilinear = bilinear
+        if not bilinear:
+            self.up = nn.ConvTranspose2d(in_ch // 2, in_ch // 2, 2, stride=2)
+
+        self.conv = double_res_conv(in_ch, out_ch, bn)
+
+    def forward(self, x1, x2):
+        if not self.bilinear:
+            x1 = self.up(x1)
+        else:
+            x1 = nn.functional.interpolate(x1, scale_factor=2, mode='bilinear', align_corners=True)
+
+        diffY = x2.size()[2] - x1.size()[2]
+        diffX = x2.size()[3] - x1.size()[3]
+
+        x1 = F.pad(x1, (diffX // 2, diffX - diffX // 2, diffY // 2, diffY - diffY // 2))
+
+        x = torch.cat([x2, x1], dim=1)
+        x = self.conv(x)
+        return x
+
+
+class outconv(nn.Module):
+    def __init__(self, in_ch, out_ch):
+        super(outconv, self).__init__()
+        self.conv = nn.Conv2d(in_ch, out_ch, 1, padding=0)
+
+    def forward(self, x):
+        x = self.conv(x)
+
+        return x
+
+
+class PostUNet(nn.Module):
+    def __init__(self, n_channels=1, scale=1):
+        super(PostUNet, self).__init__()
+
+        self.inc = inconv(n_channels, 64 // scale)
+        self.down1 = down(64 // scale, 128 // scale)
+        self.down2 = down(128 // scale, 256 // scale)
+        self.down3 = down(256 // scale, 512 // scale)
+        self.down4 = down(512 // scale, 512 // scale)
+
+        self.up1 = up(1024 // scale, 256 // scale)
+        self.up2 = up(512 // scale, 128 // scale)
+        self.up3 = up(256 // scale, 64 // scale)
+        self.up4 = up(128 // scale, 32 // scale)
+
+        self.reduce = outconv(32 // scale, 1)
+
+    def forward(self, x0):
+        x1 = self.inc(x0)
+        x2 = self.down1(x1)
+        x3 = self.down2(x2)
+        x4 = self.down3(x3)
+        x5 = self.down4(x4)
+        x = self.up1(x5, x4)
+        x = self.up2(x, x3)
+        x = self.up3(x, x2)
+        x = self.up4(x, x1)
+        x = self.reduce(x)
+        x = x[:, 0, :, :]
+        return x

mtts/models/vocoder/VocGAN/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2020 Rishikesh
+
+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.

mtts/models/vocoder/VocGAN/README.md

@@ -0,0 +1,60 @@
+# Modified VocGAN
+<noscript><a href="https://liberapay.com/rishikksh20/donate"><img alt="Donate using Liberapay" src="https://liberapay.com/assets/widgets/donate.svg"></a></noscript> <br />
+This repo implements modified version of [VocGAN: A High-Fidelity Real-time Vocoder with a Hierarchically-nested Adversarial Network](https://arxiv.org/abs/2007.15256) using Pytorch, for actual VocGAN checkout to `baseline` branch. I bit modify the VocGAN's generator and used Full-Band MelGAN's discriminator instead of VocGAN's discriminator, as in my research I found MelGAN's discriminator is very fast while training and enough powerful to train Generator to produce high fidelity voice whereas VocGAN Hierarchically-nested JCU discriminator is quite huge and extremely slows the training process.
+
+![](./assets/vocgan.JPG)
+
+Tested on Python 3.6
+```bash
+pip install -r requirements.txt
+```
+
+## Prepare Dataset
+
+- Download dataset for training. This can be any wav files with sample rate 22050Hz. (e.g. LJSpeech was used in paper)
+- preprocess: `python preprocess.py -c config/default.yaml -d [data's root path]`
+- Edit configuration `yaml` file
+
+## Train & Tensorboard
+
+- `python trainer.py -c [config yaml file] -n [name of the run]`
+  - `cp config/default.yaml config/config.yaml` and then edit `config.yaml`
+  - Write down the root path of train/validation files to 2nd/3rd line.
+  
+- `tensorboard --logdir logs/`
+
+## Notes
+1) This repo implements modified VocGAN for faster training although for true VocGAN implementation please checkout `baseline` branch, In my testing I am available to generate High-Fidelity audio in real time from Modified VocGAN.
+2) Training cost for `baseline` VocGAN's `Discriminator` is too high (2.8 sec/it on P100 with batch size 16) as compared to `Generator` (7.2 it/sec on P100 with batch size 16), so it's unfeasible for me to train this model for long time.
+3) May be we can optimizer `baseline` VocGAN's Discriminator by downsampling the audio on pre-processing stage instead of Training stage (currently I used `torchaudio.transform.Resample` as layer for downsampling the audio), this step might be speed-up overall `Discriminator` training.
+4) I trained `baseline` model for 300 epochs (with batch size 16) on LJSpeech, and quality of generated audio is similar to the MelGAN at same epoch on same dataset. Author recommend to train model till 3000 epochs which is not feasible at current training speed `(2.80 sec/it)`.
+5) I am open for any suggestion and modification on this repo.
+
+## Inference
+
+- `python inference.py -p [checkpoint path] -i [input mel path]`
+
+
+## Pretrained models
+Two pretrained model are provided. Both pretrained models are trained using modified-VocGAN structure.
+* [LJSpeech-1.1](https://keithito.com/LJ-Speech-Dataset/)  (English, single-female speaker, trained for 4000 epochs) \[ [download](https://drive.google.com/file/d/1YCXTbPtVQ3aev7KsL0Cv0kZguEMKZuQp/view?usp=sharing) \]
+* [KSS dataset](https://www.kaggle.com/bryanpark/korean-single-speaker-speech-dataset)  (Korean, single-female speaker, trained for 4500 epochs) \[ [download](https://drive.google.com/file/d/1RIqMyj9vSCcGaRgAKrVB5Duvw22N6IXJ/view?usp=sharing) \]
+* [VCTK](https://datashare.ed.ac.uk/handle/10283/3443) (English, multispeaker, trained for 3180 epochs) \[[download](https://drive.google.com/file/d/1nfD84ot7o3u2tFR7YkSp2vQWVnNJ-md_/view?usp=sharing)\]
+
+## Audio Samples
+Using pretrained models, we can reconstruct audio samples. Visit [here](https://drive.google.com/drive/folders/1QRS9BOQeOXV1aJHdz0ccoKzgfXJFbIKm?usp=sharing) to listen.
+
+
+## Results
+[WIP]
+
+## References
+- [VocGAN](https://arxiv.org/abs/2007.15256)
+- [Multi-band MelGAN](https://arxiv.org/abs/2005.05106)
+- [MelGAN](https://arxiv.org/abs/1910.06711)
+- [Pytorch implementation of melgan](https://github.com/seungwonpark/melgan)
+- [Official implementation of melgan](https://github.com/descriptinc/melgan-neurips)
+- [Multi, Full-band melgan implementation](https://github.com/rishikksh20/melgan)
+- [Nvidia's pre-processing](https://github.com/NVIDIA/tacotron2)
+- [WaveRNN](https://github.com/fatchord/WaveRNN)
+

mtts/models/vocoder/VocGAN/__init__.py

@@ -0,0 +1 @@
+from .vocgan import VocGan

mtts/models/vocoder/VocGAN/config/default.yaml

@@ -0,0 +1,60 @@
+data: # root path of train/validation data (either relative/absoulte path is ok)
+  train: 'H:\Deepsync\backup\deepsync\LJSpeech-1.1\wavs'
+  validation: 'H:\Deepsync\backup\deepsync\LJSpeech-1.1\valid'
+  mel_path: 'H:\Deepsync\backup\fastspeech\data\mels'
+  eval_path: ''
+---
+train:
+  rep_discriminator: 1
+  discriminator_train_start_steps: 100000
+  num_workers: 8
+  batch_size: 16
+  optimizer: 'adam'
+  adam:
+    lr: 0.0001
+    beta1: 0.5
+    beta2: 0.9
+---
+audio:
+  n_mel_channels: 80
+  segment_length: 16000
+  pad_short: 2000
+  filter_length: 1024
+  hop_length: 256 # WARNING: this can't be changed.
+  win_length: 1024
+  sampling_rate: 22050
+  mel_fmin: 0.0
+  mel_fmax: 8000.0
+---
+model:
+  feat_match: 10.0
+  lambda_adv: 1
+  use_subband_stft_loss: False
+  feat_loss: False
+  out_channels: 1
+  generator_ratio: [4, 4, 2, 2, 2, 2] # for 256 hop size and 22050 sample rate
+  mult: 256
+  n_residual_layers: 4
+  num_D : 3
+  ndf : 16
+  n_layers: 3
+  downsampling_factor: 4
+  disc_out: 512
+  
+stft_loss_params:
+    fft_sizes: [1024, 2048, 512]  # List of FFT size for STFT-based loss.
+    hop_sizes: [120, 240, 50]     # List of hop size for STFT-based loss
+    win_lengths: [600, 1200, 240] # List of window length for STFT-based loss.
+    window: "hann_window"         # Window function for STFT-based loss
+subband_stft_loss_params:
+  fft_sizes: [384, 683, 171]  # List of FFT size for STFT-based loss.
+  hop_sizes: [30, 60, 10]     # List of hop size for STFT-based loss
+  win_lengths: [150, 300, 60] # List of window length for STFT-based loss.
+  window: "hann_window"       # Window function for STFT-based loss
+---
+log:
+  summary_interval: 1
+  validation_interval: 5
+  save_interval: 20
+  chkpt_dir: 'chkpt'
+  log_dir: 'logs'

mtts/models/vocoder/VocGAN/datasets/dataloader.py

@@ -0,0 +1,104 @@
+import os
+import glob
+import torch
+import random
+import numpy as np
+from torch.utils.data import Dataset, DataLoader
+import torchaudio
+from utils.utils import read_wav_np
+
+
+def create_dataloader(hp, args, train):
+    dataset = MelFromDisk(hp, args, train)
+
+    if train:
+        return DataLoader(dataset=dataset, batch_size=hp.train.batch_size, shuffle=True,
+            num_workers=hp.train.num_workers, pin_memory=True, drop_last=True)
+    else:
+        return DataLoader(dataset=dataset, batch_size=1, shuffle=False,
+            num_workers=0, pin_memory=False, drop_last=False)
+
+
+class MelFromDisk(Dataset):
+    def __init__(self, hp, args, train):
+        self.hp = hp
+        self.args = args
+        self.train = train
+        self.path = hp.data.train if train else hp.data.validation
+        self.wav_list = glob.glob(os.path.join(self.path, '**', '*.wav'), recursive=True)
+        #print("Wavs path :", self.path)
+        #print(self.hp.data.mel_path)
+        #print("Length of wavelist :", len(self.wav_list))
+        self.mel_segment_length = hp.audio.segment_length // hp.audio.hop_length + 2
+        self.mapping = [i for i in range(len(self.wav_list))]
+
+    def __len__(self):
+        return len(self.wav_list)
+
+    def __getitem__(self, idx):
+
+        if self.train:
+            idx1 = idx
+            idx2 = self.mapping[idx1]
+            return self.my_getitem(idx1), self.my_getitem(idx2)
+        else:
+            return self.my_getitem(idx)
+
+    def shuffle_mapping(self):
+        random.shuffle(self.mapping)
+
+    def my_getitem(self, idx):
+        wavpath = self.wav_list[idx]
+        id = os.path.basename(wavpath).split(".")[0]
+
+        mel_path = "{}/{}.npy".format(self.hp.data.mel_path, id)
+
+        sr, audio = read_wav_np(wavpath)
+        if len(audio) < self.hp.audio.segment_length + self.hp.audio.pad_short:
+            audio = np.pad(audio, (0, self.hp.audio.segment_length + self.hp.audio.pad_short - len(audio)), \
+                    mode='constant', constant_values=0.0)
+
+        audio = torch.from_numpy(audio).unsqueeze(0)
+        # mel = torch.load(melpath).squeeze(0) # # [num_mel, T]
+
+        mel = torch.from_numpy(np.load(mel_path))
+
+        if self.train:
+            max_mel_start = mel.size(1) - self.mel_segment_length
+            mel_start = random.randint(0, max_mel_start)
+            mel_end = mel_start + self.mel_segment_length
+            mel = mel[:, mel_start:mel_end]
+
+            audio_start = mel_start * self.hp.audio.hop_length
+            audio = audio[:, audio_start:audio_start+self.hp.audio.segment_length]
+
+        audio = audio + (1/32768) * torch.randn_like(audio)
+        return mel, audio
+
+
+def collate_fn(batch):
+
+    sr = 22050
+    # perform padding and conversion to tensor
+    mels_g = [x[0][0] for x in batch]
+    audio_g = [x[0][1] for x in batch]
+
+    mels_g = torch.stack(mels_g)
+    audio_g = torch.stack(audio_g)
+
+    sub_orig_1 = torchaudio.transforms.Resample(sr, (sr // 2))(audio_g)
+    sub_orig_2 = torchaudio.transforms.Resample(sr, (sr // 4))(audio_g)
+    sub_orig_3 = torchaudio.transforms.Resample(sr, (sr // 8))(audio_g)
+    sub_orig_4 = torchaudio.transforms.Resample(sr, (sr // 16))(audio_g)
+
+    mels_d = [x[1][0] for x in batch]
+    audio_d = [x[1][1] for x in batch]
+    mels_d = torch.stack(mels_d)
+    audio_d = torch.stack(audio_d)
+    sub_orig_1_d = torchaudio.transforms.Resample(sr, (sr // 2))(audio_d)
+    sub_orig_2_d = torchaudio.transforms.Resample(sr, (sr // 4))(audio_d)
+    sub_orig_3_d = torchaudio.transforms.Resample(sr, (sr // 8))(audio_d)
+    sub_orig_4_d = torchaudio.transforms.Resample(sr, (sr // 16))(audio_d)
+
+    return [mels_g, audio_g, sub_orig_1, sub_orig_2, sub_orig_3, sub_orig_4],\
+           [mels_d, audio_d, sub_orig_1_d, sub_orig_2_d, sub_orig_3_d, sub_orig_4_d]

mtts/models/vocoder/VocGAN/denoiser.py

@@ -0,0 +1,38 @@
+import sys
+import torch
+from .utils.stft import STFT
+
+
+class Denoiser(torch.nn.Module):
+    """ Removes model bias from audio produced with waveglow """
+
+    def __init__(self, melgan, filter_length=1024, n_overlap=4,
+                 win_length=1024, mode='zeros',device='cpu'):
+        super(Denoiser, self).__init__()
+        self.stft = STFT(filter_length=filter_length,
+                         hop_length=int(filter_length/n_overlap),
+                         win_length=win_length,device=device).to(device)
+        if mode == 'zeros':
+            mel_input = torch.zeros(
+                (1, 80, 88)).to(device)
+        elif mode == 'normal':
+            mel_input = torch.randn(
+                (1, 80, 88)).to(device)
+        else:
+            raise Exception("Mode {} if not supported".format(mode))
+
+        with torch.no_grad():
+            bias_audio = melgan.inference(mel_input).float() # [B, 1, T]
+
+            bias_spec, _ = self.stft.transform(bias_audio.squeeze(0))
+
+        self.register_buffer('bias_spec', bias_spec[:, :, 0][:, :, None])
+        self.device = device
+        self.to(device)
+
+    def forward(self, audio, strength=0.1):
+        audio_spec, audio_angles = self.stft.transform(audio.to(self.device).float())
+        audio_spec_denoised = audio_spec.to(self.device) - self.bias_spec * strength
+        audio_spec_denoised = torch.clamp(audio_spec_denoised, 0.0)
+        audio_denoised = self.stft.inverse(audio_spec_denoised, audio_angles.to(self.device))
+        return audio_denoised

mtts/models/vocoder/VocGAN/download_utils.py

@@ -0,0 +1,175 @@
+import hashlib
+import logging
+import os
+import tarfile
+import threading
+import urllib
+import urllib.request
+import zipfile
+from queue import Queue
+from typing import Any, Iterable, List, Optional
+import random
+
+import torch
+from tqdm import tqdm
+
+
+def stream_url(url: str,
+               start_byte: Optional[int] = None,
+               block_size: int = 32 * 1024,
+               progress_bar: bool = True) -> Iterable:
+    """Stream url by chunk
+    Args:
+        url (str): Url.
+        start_byte (int, optional): Start streaming at that point (Default: ``None``).
+        block_size (int, optional): Size of chunks to stream (Default: ``32 * 1024``).
+        progress_bar (bool, optional): Display a progress bar (Default: ``True``).
+    """
+
+    # If we already have the whole file, there is no need to download it again
+    req = urllib.request.Request(url, method="HEAD")
+    url_size = int(urllib.request.urlopen(req).info().get("Content-Length", -1))
+    if url_size == start_byte:
+        return
+
+    req = urllib.request.Request(url)
+    if start_byte:
+        req.headers["Range"] = "bytes={}-".format(start_byte)
+
+    with urllib.request.urlopen(req) as upointer, tqdm(
+            unit="B",
+            unit_scale=True,
+            unit_divisor=1024,
+            total=url_size,
+            disable=not progress_bar,
+    ) as pbar:
+
+        num_bytes = 0
+        while True:
+            chunk = upointer.read(block_size)
+            if not chunk:
+                break
+            yield chunk
+            num_bytes += len(chunk)
+            pbar.update(len(chunk))
+def validate_file(file_obj: Any, hash_value: str, hash_type: str = "sha256") -> bool:
+    """Validate a given file object with its hash.
+    Args:
+        file_obj: File object to read from.
+        hash_value (str): Hash for url.
+        hash_type (str, optional): Hash type, among "sha256" and "md5" (Default: ``"sha256"``).
+    Returns:
+        bool: return True if its a valid file, else False.
+    """
+
+    if hash_type == "sha256":
+        hash_func = hashlib.sha256()
+    elif hash_type == "md5":
+        hash_func = hashlib.md5()
+    else:
+        raise ValueError
+
+    while True:
+        # Read by chunk to avoid filling memory
+        chunk = file_obj.read(1024 ** 2)
+        if not chunk:
+            break
+        hash_func.update(chunk)
+
+    return hash_func.hexdigest() == hash_value
+
+
+
+def download_url(url: str,
+                 download_folder: str,
+                 filename: Optional[str] = None,
+                 hash_value: Optional[str] = None,
+                 hash_type: str = "sha256",
+                 progress_bar: bool = True,
+                 resume: bool = False) -> None:
+    """Download file to disk.
+    Args:
+        url (str): Url.
+        download_folder (str): Folder to download file.
+        filename (str, optional): Name of downloaded file. If None, it is inferred from the url (Default: ``None``).
+        hash_value (str, optional): Hash for url (Default: ``None``).
+        hash_type (str, optional): Hash type, among "sha256" and "md5" (Default: ``"sha256"``).
+        progress_bar (bool, optional): Display a progress bar (Default: ``True``).
+        resume (bool, optional): Enable resuming download (Default: ``False``).
+    """
+
+    req = urllib.request.Request(url, method="HEAD")
+    req_info = urllib.request.urlopen(req).info()
+
+    # Detect filename
+    filename = filename or req_info.get_filename() or os.path.basename(url)
+    filepath = os.path.join(download_folder, filename)
+    if resume and os.path.exists(filepath):
+        mode = "ab"
+        local_size: Optional[int] = os.path.getsize(filepath)
+
+    elif not resume and os.path.exists(filepath):
+        raise RuntimeError(
+            "{} already exists. Delete the file manually and retry.".format(filepath)
+        )
+    else:
+        mode = "wb"
+        local_size = None
+
+    if hash_value and local_size == int(req_info.get("Content-Length", -1)):
+        with open(filepath, "rb") as file_obj:
+            if validate_file(file_obj, hash_value, hash_type):
+                return
+        raise RuntimeError(
+            "The hash of {} does not match. Delete the file manually and retry.".format(
+                filepath
+            )
+        )
+
+    with open(filepath, mode) as fpointer:
+        for chunk in stream_url(url, start_byte=local_size, progress_bar=progress_bar):
+            fpointer.write(chunk)
+
+    with open(filepath, "rb") as file_obj:
+        if hash_value and not validate_file(file_obj, hash_value, hash_type):
+            raise RuntimeError(
+                "The hash of {} does not match. Delete the file manually and retry.".format(
+                    filepath
+                )
+            )
+    #return filepath
+            
+            
+def download_checkpoint():
+    url = 'https://zenodo.org/record/4625672/files/checkpoint_500000.pth'
+    os.makedirs('./checkpoint/',exist_ok=True)
+    return download_url(url,'./checkpoint/',resume=True,hash_value='14002c23879f6b5d0cd987f3c3e1a160',hash_type='md5')
+
+def download_waveglow(device):
+    
+    os.makedirs('./waveglow/',exist_ok=True)
+    
+    try:
+         waveglow = torch.hub.load('./waveglow/DeepLearningExamples-torchhub/', 'nvidia_waveglow',source='local')
+    except:
+        if random.randint(0,1)==0:
+            download_url('https://hub.fastgit.org/nvidia/DeepLearningExamples/archive/torchhub.zip','./waveglow',
+                         hash_type='md5',hash_value='27ef24b9c4a2ce6c26f26998aee26f44',
+                         resume=True)
+        else:
+            download_url('https://github.com/nvidia/DeepLearningExamples/archive/torchhub.zip','./waveglow',
+                         hash_type='md5',hash_value='27ef24b9c4a2ce6c26f26998aee26f44',
+                         resume=True)
+        os.system('unzip ./waveglow/DeepLearningExamples-torchhub.zip -d ./waveglow/')
+        waveglow = torch.hub.load('./waveglow/DeepLearningExamples-torchhub/', 'nvidia_waveglow',source='local')
+        
+    
+   
+    waveglow = waveglow.remove_weightnorm(waveglow)
+    waveglow.eval()
+    for m in waveglow.modules():
+        if 'Conv' in str(type(m)):
+            setattr(m, 'padding_mode', 'zeros')
+    waveglow.to(device)
+    return waveglow
+#download_url('https://github.com/nvidia/DeepLearningExamples/archive/torchhub.zip','./waveglow',resume=True)

mtts/models/vocoder/VocGAN/export_torchscript.py

@@ -0,0 +1,63 @@
+import torch
+import argparse
+from scipy.io.wavfile import write
+import numpy as np
+from model.generator import ModifiedGenerator
+from utils.hparams import HParam, load_hparam_str
+from denoiser import Denoiser
+
+MAX_WAV_VALUE = 32768.0
+
+
+def main(args):
+    checkpoint = torch.load(args.checkpoint_path)
+    if args.config is not None:
+        hp = HParam(args.config)
+    else:
+        hp = load_hparam_str(checkpoint['hp_str'])
+
+    model = ModifiedGenerator(hp.audio.n_mel_channels, hp.model.n_residual_layers,
+                        ratios=hp.model.generator_ratio, mult = hp.model.mult,
+                        out_band = hp.model.out_channels).cuda()
+    model.load_state_dict(checkpoint['model_g'])
+    model.eval(inference=True)
+
+    with torch.no_grad():
+        mel = torch.from_numpy(np.load(args.input))
+        if len(mel.shape) == 2:
+            mel = mel.unsqueeze(0)
+        mel = mel.cuda()
+        zero = torch.full((1, 80, 10), -11.5129).to(mel.device)
+        mel = torch.cat((mel, zero), dim=2)
+        vocgan_trace = torch.jit.trace(model, mel)
+        vocgan_trace.save("{}/vocgan_ex_female_en_{}_{}.pt".format(args.out, checkpoint['githash'], checkpoint['epoch']))
+        # audio = model(mel)
+
+        # audio = audio.squeeze(0)  # collapse all dimension except time axis
+        # if args.d:
+        #     denoiser = Denoiser(model).cuda()
+        #     audio = denoiser(audio, 0.01)
+        # audio = audio.squeeze()
+        # audio = audio[:-(hp.audio.hop_length*10)]
+        # audio = MAX_WAV_VALUE * audio
+        # audio = audio.clamp(min=-MAX_WAV_VALUE, max=MAX_WAV_VALUE-1)
+        # audio = audio.short()
+        # audio = audio.cpu().detach().numpy()
+        #
+        # out_path = args.input.replace('.npy', '_reconstructed_epoch%04d.wav' % checkpoint['epoch'])
+        # write(out_path, hp.audio.sampling_rate, audio)
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('-c', '--config', type=str, default=None,
+                        help="yaml file for config. will use hp_str from checkpoint if not given.")
+    parser.add_argument('-p', '--checkpoint_path', type=str, required=True,
+                        help="path of checkpoint pt file for evaluation")
+    parser.add_argument('-i', '--input', type=str, required=True,
+                        help="directory of mel-spectrograms to invert into raw audio. ")
+    parser.add_argument('-o', '--out', type=str, required=True,
+                        help="path of output pt file")
+    args = parser.parse_args()
+
+    main(args)

mtts/models/vocoder/VocGAN/inference.py

@@ -0,0 +1,61 @@
+import os
+import glob
+import tqdm
+import torch
+import argparse
+from scipy.io.wavfile import write
+import numpy as np
+from model.generator import ModifiedGenerator
+from utils.hparams import HParam, load_hparam_str
+from denoiser import Denoiser
+
+MAX_WAV_VALUE = 32768.0
+
+
+def main(args):
+    checkpoint = torch.load(args.checkpoint_path)
+    if args.config is not None:
+        hp = HParam(args.config)
+    else:
+        hp = load_hparam_str(checkpoint['hp_str'])
+
+    model = ModifiedGenerator(hp.audio.n_mel_channels, hp.model.n_residual_layers,
+                        ratios=hp.model.generator_ratio, mult = hp.model.mult,
+                        out_band = hp.model.out_channels).cuda()
+    model.load_state_dict(checkpoint['model_g'])
+    model.eval(inference=True)
+
+    with torch.no_grad():
+        mel = torch.from_numpy(np.load(args.input))
+        if len(mel.shape) == 2:
+            mel = mel.unsqueeze(0)
+        mel = mel.cuda()
+        audio = model.inference(mel)
+
+        audio = audio.squeeze(0)  # collapse all dimension except time axis
+        if args.d:
+            denoiser = Denoiser(model).cuda()
+            audio = denoiser(audio, 0.01)
+        audio = audio.squeeze()
+        audio = audio[:-(hp.audio.hop_length*10)]
+        audio = MAX_WAV_VALUE * audio
+        audio = audio.clamp(min=-MAX_WAV_VALUE, max=MAX_WAV_VALUE-1)
+        audio = audio.short()
+        audio = audio.cpu().detach().numpy()
+
+        out_path = args.input.replace('.npy', '_reconstructed_epoch%04d.wav' % checkpoint['epoch'])
+        write(out_path, hp.audio.sampling_rate, audio)
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('-c', '--config', type=str, default=None,
+                        help="yaml file for config. will use hp_str from checkpoint if not given.")
+    parser.add_argument('-p', '--checkpoint_path', type=str, required=True,
+                        help="path of checkpoint pt file for evaluation")
+    parser.add_argument('-i', '--input', type=str, required=True,
+                        help="directory of mel-spectrograms to invert into raw audio. ")
+    parser.add_argument('-d', action='store_true', help="denoising ")
+    args = parser.parse_args()
+
+    main(args)

mtts/models/vocoder/VocGAN/model/discriminator.py

@@ -0,0 +1,144 @@
+import torch
+import torch.nn as nn
+
+
+class Discriminator(nn.Module):
+    def __init__(self, ndf = 16, n_layers = 3, downsampling_factor = 4, disc_out = 512):
+        super(Discriminator, self).__init__()
+        discriminator = nn.ModuleDict()
+        discriminator["layer_0"] = nn.Sequential(
+            nn.ReflectionPad1d(7),
+            nn.utils.weight_norm(nn.Conv1d(1, ndf, kernel_size=15, stride=1)),
+            nn.LeakyReLU(0.2, True),
+        )
+
+        nf = ndf
+        stride = downsampling_factor
+        for n in range(1, n_layers + 1):
+            nf_prev = nf
+            nf = min(nf * stride, disc_out)
+
+            discriminator["layer_%d" % n] = nn.Sequential(
+                nn.utils.weight_norm(nn.Conv1d(
+                    nf_prev,
+                    nf,
+                    kernel_size=stride * 10 + 1,
+                    stride=stride,
+                    padding=stride * 5,
+                    groups=nf_prev // 4,
+                )),
+                nn.LeakyReLU(0.2, True),
+            )
+        nf = min(nf * 2, disc_out)
+        discriminator["layer_%d" % (n_layers + 1)] = nn.Sequential(
+            nn.utils.weight_norm(nn.Conv1d(nf, disc_out, kernel_size=5, stride=1, padding=2)),
+            nn.LeakyReLU(0.2, True),
+        )
+
+        discriminator["layer_%d" % (n_layers + 2)] = nn.utils.weight_norm(nn.Conv1d(
+            nf, 1, kernel_size=3, stride=1, padding=1
+        ))
+        self.discriminator = discriminator
+
+    def forward(self, x):
+        '''
+            returns: (list of 6 features, discriminator score)
+            we directly predict score without last sigmoid function
+            since we're using Least Squares GAN (https://arxiv.org/abs/1611.04076)
+        '''
+        features = list()
+        for key, module in self.discriminator.items():
+            x = module(x)
+            features.append(x)
+        return features[:-1], features[-1]
+
+# JCU Discriminator
+class JCU_Discriminator(nn.Module):
+    def __init__(self):
+        super(JCU_Discriminator, self).__init__()
+        self.mel_conv = nn.Sequential(
+            nn.ReflectionPad1d(3),
+            nn.utils.weight_norm(nn.Conv1d(80, 128, kernel_size=2, stride=1)),
+            nn.LeakyReLU(0.2, True),
+        )
+        x_conv = [nn.ReflectionPad1d(7),
+            nn.utils.weight_norm(nn.Conv1d(1, 16, kernel_size=7, stride=1)),
+            nn.LeakyReLU(0.2, True),
+            ]
+        x_conv += [
+                        nn.utils.weight_norm(nn.Conv1d(
+                                        16,
+                                        64,
+                                        kernel_size=41,
+                                        stride=4,
+                                        padding=4 * 5,
+                                        groups=16 // 4,
+                                    )
+                                    ),
+                nn.LeakyReLU(0.2),
+            ]
+        x_conv += [
+            nn.utils.weight_norm(nn.Conv1d(
+                64,
+                128,
+                kernel_size=21,
+                stride=2,
+                padding=2 * 5,
+                groups=64 // 4,
+            )
+            ),
+            nn.LeakyReLU(0.2),
+        ]
+        self.x_conv = nn.Sequential(*x_conv)
+        self.mel_conv2 = nn.Sequential(
+            nn.utils.weight_norm(nn.Conv1d(128, 128, kernel_size=5, stride=1, padding=2)),
+            nn.LeakyReLU(0.2, True),
+        )
+        self.mel_conv3 = nn.utils.weight_norm(nn.Conv1d(
+            128, 1, kernel_size=3, stride=1, padding=1
+        ))
+
+        self.x_conv2 = nn.Sequential(
+            nn.utils.weight_norm(nn.Conv1d(128, 128, kernel_size=5, stride=1, padding=2)),
+            nn.LeakyReLU(0.2, True),
+        )
+        self.x_conv3 = nn.utils.weight_norm(nn.Conv1d(
+            128, 1, kernel_size=3, stride=1, padding=1
+        ))
+
+    def forward(self, x, mel):
+        out = self.mel_conv(mel)
+        out1 = self.x_conv(x)
+        out = torch.cat([out, out1], dim=2)
+        out = self.mel_conv2(out)
+        cond_out = self.mel_conv3(out)
+        out1 = self.x_conv2(out1)
+        uncond_out = self.x_conv3(out1)
+        return uncond_out, cond_out
+
+    
+if __name__ == '__main__':
+    model = Discriminator()
+    '''
+    Length of features :  5
+    Length of score :  3
+    torch.Size([3, 16, 25600])
+    torch.Size([3, 64, 6400])
+    torch.Size([3, 256, 1600])
+    torch.Size([3, 512, 400])
+    torch.Size([3, 512, 400])
+    torch.Size([3, 1, 400]) -> score
+    '''
+
+    x = torch.randn(3, 1, 25600)
+    print(x.shape)
+
+    features, score = model(x)
+    print("Length of features : ", len(features))
+    print("Length of score : ", len(score))
+    for feat in features:
+        print(feat.shape)
+    print(score.shape)
+
+    pytorch_total_params = sum(p.numel() for p in model.parameters() if p.requires_grad)
+    print(pytorch_total_params)

mtts/models/vocoder/VocGAN/model/generator.py

@@ -0,0 +1,422 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from ..utils.utils import weights_init
+from .res_stack import ResStack
+# from res_stack import ResStack
+
+MAX_WAV_VALUE = 32768.0
+
+
+class VOCGenerator(nn.Module):
+    def __init__(self, mel_channel, n_residual_layers, ratios=[4, 4, 2, 2, 2, 2], mult=256, out_band=1):
+        super(VOCGenerator, self).__init__()
+        self.mel_channel = mel_channel
+
+        self.start = nn.Sequential(
+            nn.ReflectionPad1d(3),
+            nn.utils.weight_norm(nn.Conv1d(mel_channel, mult * 2, kernel_size=7, stride=1))
+        )
+        r = ratios[0]
+        self.upsample_1 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.utils.weight_norm(nn.ConvTranspose1d(mult * 2, mult,
+                                                    kernel_size=r * 2, stride=r,
+                                                    padding=r // 2 + r % 2,
+                                                    output_padding=r % 2)
+                                 )
+        )
+        self.res_stack_1 = nn.Sequential(*[ResStack(mult, dilation=3 ** j) for j in range(n_residual_layers)])
+
+        r = ratios[1]
+        mult = mult // 2
+        self.upsample_2 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.utils.weight_norm(nn.ConvTranspose1d(mult * 2, mult,
+                                                    kernel_size=r * 2, stride=r,
+                                                    padding=r // 2 + r % 2,
+                                                    output_padding=r % 2)
+                                 )
+        )
+        self.res_stack_2 = nn.Sequential(*[ResStack(mult, dilation=3 ** j) for j in range(n_residual_layers)])
+
+
+        self.sub_out_1 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.ReflectionPad1d(3),
+            nn.utils.weight_norm(nn.Conv1d(mult, out_band, kernel_size=7, stride=1)),
+            nn.Tanh(),
+        )
+
+        r = ratios[2]
+        mult = mult // 2
+        self.upsample_3 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.utils.weight_norm(nn.ConvTranspose1d(mult * 2, mult,
+                                                    kernel_size=r * 2, stride=r,
+                                                    padding=r // 2 + r % 2,
+                                                    output_padding=r % 2)
+                                 )
+        )
+
+        self.skip_upsample_1 = nn.utils.weight_norm(nn.ConvTranspose1d(mel_channel, mult,
+                                                                       kernel_size=64, stride=32,
+                                                                       padding=16,
+                                                                       output_padding=0)
+                                                    )
+        self.res_stack_3 = nn.Sequential(*[ResStack(mult, dilation=3 ** j) for j in range(n_residual_layers)])
+
+        self.sub_out_2 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.ReflectionPad1d(3),
+            nn.utils.weight_norm(nn.Conv1d(mult, out_band, kernel_size=7, stride=1)),
+            nn.Tanh(),
+        )
+
+        r = ratios[3]
+        mult = mult // 2
+        self.upsample_4 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.utils.weight_norm(nn.ConvTranspose1d(mult * 2, mult,
+                                                    kernel_size=r * 2, stride=r,
+                                                    padding=r // 2 + r % 2,
+                                                    output_padding=r % 2)
+                                 )
+        )
+
+        self.skip_upsample_2 = nn.utils.weight_norm(nn.ConvTranspose1d(mel_channel, mult,
+                                                                       kernel_size=128, stride=64,
+                                                                       padding=32,
+                                                                       output_padding=0)
+                                                    )
+        self.res_stack_4 = nn.Sequential(*[ResStack(mult, dilation=3 ** j) for j in range(n_residual_layers)])
+
+        self.sub_out_3 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.ReflectionPad1d(3),
+            nn.utils.weight_norm(nn.Conv1d(mult, out_band, kernel_size=7, stride=1)),
+            nn.Tanh(),
+        )
+
+        r = ratios[4]
+        mult = mult // 2
+        self.upsample_5 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.utils.weight_norm(nn.ConvTranspose1d(mult * 2, mult,
+                                                    kernel_size=r * 2, stride=r,
+                                                    padding=r // 2 + r % 2,
+                                                    output_padding=r % 2)
+                                 )
+        )
+
+        self.skip_upsample_3 = nn.utils.weight_norm(nn.ConvTranspose1d(mel_channel, mult,
+                                                                       kernel_size=256, stride=128,
+                                                                       padding=64,
+                                                                       output_padding=0)
+                                                    )
+        self.res_stack_5 = nn.Sequential(*[ResStack(mult, dilation=3 ** j) for j in range(n_residual_layers)])
+
+        self.sub_out_4 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.ReflectionPad1d(3),
+            nn.utils.weight_norm(nn.Conv1d(mult, out_band, kernel_size=7, stride=1)),
+            nn.Tanh(),
+        )
+
+        r = ratios[5]
+        mult = mult // 2
+        self.upsample_6 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.utils.weight_norm(nn.ConvTranspose1d(mult * 2, mult,
+                                                    kernel_size=r * 2, stride=r,
+                                                    padding=r // 2 + r % 2,
+                                                    output_padding=r % 2)
+                                 )
+        )
+
+        self.skip_upsample_4 = nn.utils.weight_norm(nn.ConvTranspose1d(mel_channel, mult,
+                                                                       kernel_size=512, stride=256,
+                                                                       padding=128,
+                                                                       output_padding=0)
+                                                    )
+        self.res_stack_6 = nn.Sequential(*[ResStack(mult, dilation=3 ** j) for j in range(n_residual_layers)])
+
+        self.out = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.ReflectionPad1d(3),
+            nn.utils.weight_norm(nn.Conv1d(mult, out_band, kernel_size=7, stride=1)),
+            nn.Tanh(),
+        )
+        self.apply(weights_init)
+
+    def forward(self, mel):
+        mel = (mel + 5.0) / 5.0 # roughly normalize spectrogram
+        # Mel Shape [B, num_mels, T] -> torch.Size([3, 80, 10])
+        x = self.start(mel)  # [B, dim*2, T] -> torch.Size([3, 512, 10])
+
+        x = self.upsample_1(x)
+        x = self.res_stack_1(x)  # [B, dim, T*4] -> torch.Size([3, 256, 40])
+
+        x = self.upsample_2(x)
+        x = self.res_stack_2(x)  # [B, dim/2, T*16] -> torch.Size([3, 128, 160])
+        out1 = self.sub_out_1(x)  # [B, 1, T*16] -> torch.Size([3, 1, 160])
+
+        x = self.upsample_3(x)
+        x = x + self.skip_upsample_1(mel)
+        x = self.res_stack_3(x)  # [B, dim/4, T*32] -> torch.Size([3, 64, 320])
+        out2 = self.sub_out_2(x)  # [B, 1, T*32] -> torch.Size([3, 1, 320])
+
+        x = self.upsample_4(x)
+        x = x + self.skip_upsample_2(mel)
+        x = self.res_stack_4(x)  # [B, dim/8, T*64] -> torch.Size([3, 32, 640])
+        out3 = self.sub_out_3(x)  # [B, 1, T*64] -> torch.Size([3, 1, 640])
+
+        x = self.upsample_5(x)
+        x = x + self.skip_upsample_3(mel)
+        x = self.res_stack_5(x)  # [B, dim/16, T*128] -> torch.Size([3, 16, 1280])
+        out4 = self.sub_out_4(x)  # [B, 1, T*128] -> torch.Size([3, 1, 1280])
+
+        x = self.upsample_6(x)
+        x = x + self.skip_upsample_4(mel)
+        x = self.res_stack_6(x)  # [B, dim/32, T*256] -> torch.Size([3, 8, 2560])
+
+        out = self.out(x)  # [B, 1, T*256] -> torch.Size([3, 1, 2560])
+
+        return out1, out2, out3, out4, out
+
+    def inference(self, mel):
+        hop_length = 256
+        # pad input mel with zeros to cut artifact
+        # see https://github.com/seungwonpark/melgan/issues/8
+        zero = torch.full((1, self.mel_channel, 10), -11.5129).to(mel.device)
+        mel = torch.cat((mel, zero), dim=2)
+
+        _, _, _, _, audio = self.forward(mel)
+        return audio
+
+
+class ModifiedGenerator(nn.Module):
+    def __init__(self, mel_channel, n_residual_layers, ratios=[4, 4, 2, 2, 2, 2], mult=256, out_band=1):
+        super(ModifiedGenerator, self).__init__()
+        self.mel_channel = mel_channel
+
+        self.start = nn.Sequential(
+            nn.ReflectionPad1d(3),
+            nn.utils.weight_norm(nn.Conv1d(mel_channel, mult * 2, kernel_size=7, stride=1))
+        )
+        r = ratios[0]
+        self.upsample_1 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.utils.weight_norm(nn.ConvTranspose1d(mult * 2, mult,
+                                                    kernel_size=r * 2, stride=r,
+                                                    padding=r // 2 + r % 2,
+                                                    output_padding=r % 2)
+                                 )
+        )
+        self.res_stack_1 = nn.Sequential(*[ResStack(mult, dilation=3 ** j) for j in range(n_residual_layers)])
+
+        r = ratios[1]
+        mult = mult // 2
+        self.upsample_2 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.utils.weight_norm(nn.ConvTranspose1d(mult * 2, mult,
+                                                    kernel_size=r * 2, stride=r,
+                                                    padding=r // 2 + r % 2,
+                                                    output_padding=r % 2)
+                                 )
+        )
+        self.res_stack_2 = nn.Sequential(*[ResStack(mult, dilation=3 ** j) for j in range(n_residual_layers)])
+
+        r = ratios[2]
+        mult = mult // 2
+        self.upsample_3 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.utils.weight_norm(nn.ConvTranspose1d(mult * 2, mult,
+                                                    kernel_size=r * 2, stride=r,
+                                                    padding=r // 2 + r % 2,
+                                                    output_padding=r % 2)
+                                 )
+        )
+
+        self.skip_upsample_1 = nn.utils.weight_norm(nn.ConvTranspose1d(mel_channel, mult,
+                                                                       kernel_size=64, stride=32,
+                                                                       padding=16,
+                                                                       output_padding=0)
+                                                    )
+        self.res_stack_3 = nn.Sequential(*[ResStack(mult, dilation=3 ** j) for j in range(n_residual_layers)])
+
+
+
+        r = ratios[3]
+        mult = mult // 2
+        self.upsample_4 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.utils.weight_norm(nn.ConvTranspose1d(mult * 2, mult,
+                                                    kernel_size=r * 2, stride=r,
+                                                    padding=r // 2 + r % 2,
+                                                    output_padding=r % 2)
+                                 )
+        )
+
+        self.skip_upsample_2 = nn.utils.weight_norm(nn.ConvTranspose1d(mel_channel, mult,
+                                                                       kernel_size=128, stride=64,
+                                                                       padding=32,
+                                                                       output_padding=0)
+                                                    )
+        self.res_stack_4 = nn.Sequential(*[ResStack(mult, dilation=3 ** j) for j in range(n_residual_layers)])
+
+
+        r = ratios[4]
+        mult = mult // 2
+        self.upsample_5 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.utils.weight_norm(nn.ConvTranspose1d(mult * 2, mult,
+                                                    kernel_size=r * 2, stride=r,
+                                                    padding=r // 2 + r % 2,
+                                                    output_padding=r % 2)
+                                 )
+        )
+
+        self.skip_upsample_3 = nn.utils.weight_norm(nn.ConvTranspose1d(mel_channel, mult,
+                                                                       kernel_size=256, stride=128,
+                                                                       padding=64,
+                                                                       output_padding=0)
+                                                    )
+        self.res_stack_5 = nn.Sequential(*[ResStack(mult, dilation=3 ** j) for j in range(n_residual_layers)])
+
+
+        r = ratios[5]
+        mult = mult // 2
+        self.upsample_6 = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.utils.weight_norm(nn.ConvTranspose1d(mult * 2, mult,
+                                                    kernel_size=r * 2, stride=r,
+                                                    padding=r // 2 + r % 2,
+                                                    output_padding=r % 2)
+                                 )
+        )
+
+        self.skip_upsample_4 = nn.utils.weight_norm(nn.ConvTranspose1d(mel_channel, mult,
+                                                                       kernel_size=512, stride=256,
+                                                                       padding=128,
+                                                                       output_padding=0)
+                                                    )
+        self.res_stack_6 = nn.Sequential(*[ResStack(mult, dilation=3 ** j) for j in range(n_residual_layers)])
+
+        self.out = nn.Sequential(
+            nn.LeakyReLU(0.2),
+            nn.ReflectionPad1d(3),
+            nn.utils.weight_norm(nn.Conv1d(mult, out_band, kernel_size=7, stride=1)),
+            nn.Tanh(),
+        )
+        self.apply(weights_init)
+
+    def forward(self, mel):
+        mel = (mel + 5.0) / 5.0  # roughly normalize spectrogram
+        # Mel Shape [B, num_mels, T] -> torch.Size([3, 80, 10])
+        x = self.start(mel)  # [B, dim*2, T] -> torch.Size([3, 512, 10])
+
+        x = self.upsample_1(x)
+        x = self.res_stack_1(x)  # [B, dim, T*4] -> torch.Size([3, 256, 40])
+
+        x = self.upsample_2(x)
+        x = self.res_stack_2(x)  # [B, dim/2, T*16] -> torch.Size([3, 128, 160])
+        # out1 = self.sub_out_1(x)  # [B, 1, T*16] -> torch.Size([3, 1, 160])
+
+        x = self.upsample_3(x)
+        x = x + self.skip_upsample_1(mel)
+        x = self.res_stack_3(x)  # [B, dim/4, T*32] -> torch.Size([3, 64, 320])
+        # out2 = self.sub_out_2(x)  # [B, 1, T*32] -> torch.Size([3, 1, 320])
+
+        x = self.upsample_4(x)
+        x = x + self.skip_upsample_2(mel)
+        x = self.res_stack_4(x)  # [B, dim/8, T*64] -> torch.Size([3, 32, 640])
+        # out3 = self.sub_out_3(x)  # [B, 1, T*64] -> torch.Size([3, 1, 640])
+
+        x = self.upsample_5(x)
+        x = x + self.skip_upsample_3(mel)
+        x = self.res_stack_5(x)  # [B, dim/16, T*128] -> torch.Size([3, 16, 1280])
+        # out4 = self.sub_out_4(x)  # [B, 1, T*128] -> torch.Size([3, 1, 1280])
+
+        x = self.upsample_6(x)
+        x = x + self.skip_upsample_4(mel)
+        x = self.res_stack_6(x)  # [B, dim/32, T*256] -> torch.Size([3, 8, 2560])
+
+        out = self.out(x)  # [B, 1, T*256] -> torch.Size([3, 1, 2560])
+
+        return out #out1, out2, out3, out4, out
+
+    def eval(self, inference=False):
+        super(ModifiedGenerator, self).eval()
+
+        # don't remove weight norm while validation in training loop
+        if inference:
+            self.remove_weight_norm()
+
+    # def remove_weight_norm(self):
+    #     for idx, layer in enumerate(self.generator):
+    #         if len(layer.state_dict()) != 0:
+    #             try:
+    #                 nn.utils.remove_weight_norm(layer)
+    #             except:
+    #                 layer.remove_weight_norm()
+
+    def remove_weight_norm(self):
+        """Remove weight normalization module from all of the layers."""
+
+        def _remove_weight_norm(m):
+            try:
+                torch.nn.utils.remove_weight_norm(m)
+            except ValueError:  # this module didn't have weight norm
+                return
+
+        self.apply(_remove_weight_norm)
+
+    def apply_weight_norm(self):
+        """Apply weight normalization module from all of the layers."""
+
+        def _apply_weight_norm(m):
+            if isinstance(m, torch.nn.Conv1d) or isinstance(m, torch.nn.ConvTranspose1d):
+                torch.nn.utils.weight_norm(m)
+
+        self.apply(_apply_weight_norm)
+
+
+    def inference(self, mel):
+        hop_length = 256
+        # pad input mel with zeros to cut artifact
+        # see https://github.com/seungwonpark/melgan/issues/8
+        zero = torch.full((1, self.mel_channel, 10), -11.5129).to(mel.device)
+        mel = torch.cat((mel, zero), dim=2)
+
+        audio = self.forward(mel)
+        return audio
+
+
+
+
+'''
+    to run this, fix 
+    from . import ResStack
+    into
+    from res_stack import ResStack
+'''
+if __name__ == '__main__':
+    '''
+    torch.Size([3, 80, 10])
+    torch.Size([3, 1, 2560])
+    4715698
+    '''
+    model = VOCGenerator(80, 4)
+
+    x = torch.randn(3, 80, 10)  # (B, channels, T).
+    print(x.shape)
+
+    out1, out2, out3, out4, out = model(x)  # (B, 1, T ** prod(upsample_scales)
+    assert out.shape == torch.Size([3, 1, 2560])  # For normal melgan torch.Size([3, 1, 2560])
+    assert out4.shape == torch.Size([3, 1, 1280])
+    assert out3.shape == torch.Size([3, 1, 640])
+    assert out2.shape == torch.Size([3, 1, 320])
+    assert out1.shape == torch.Size([3, 1, 160])
+    pytorch_total_params = sum(p.numel() for p in model.parameters() if p.requires_grad)
+    print(pytorch_total_params)