init

README.md

-# 2023Thesis
-
+见两部分代码文件夹中的Readme文件说明
\ No newline at end of file

歌声合成/LICENSE

+MIT License
+
+Copyright (c) 2021 Jinglin Liu
+
+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.
\ No newline at end of file

歌声合成/README.md

+# DiffSinger: Singing Voice Synthesis via Shallow Diffusion Mechanism
+[![arXiv](https://img.shields.io/badge/arXiv-Paper-<COLOR>.svg)](https://arxiv.org/abs/2105.02446)
+[![GitHub Stars](https://img.shields.io/github/stars/MoonInTheRiver/DiffSinger?style=social)](https://github.com/MoonInTheRiver/DiffSinger)
+[![downloads](https://img.shields.io/github/downloads/MoonInTheRiver/DiffSinger/total.svg)](https://github.com/MoonInTheRiver/DiffSinger/releases)
+[![Hugging Face](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-blue?label=TTSDemo)](https://huggingface.co/spaces/NATSpeech/DiffSpeech) 
+[![Hugging Face](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-blue?label=SVSDemo)](https://huggingface.co/spaces/Silentlin/DiffSinger)
+
+
+This repository is the official PyTorch implementation of our AAAI-2022 [paper](https://arxiv.org/abs/2105.02446), in which we propose DiffSinger (for Singing-Voice-Synthesis) and DiffSpeech (for Text-to-Speech).
+ 
+
+:tada: :tada: :tada: **Updates**:
+ - Sep.11, 2022: :electric_plug: [DiffSinger-PN](docs/README-SVS-opencpop-pndm.md). Add plug-in [PNDM](https://arxiv.org/abs/2202.09778), ICLR 2022 in our laboratory, to accelerate DiffSinger freely.
+ - Jul.27, 2022: Update documents for [SVS](docs/README-SVS.md). Add easy inference [A](docs/README-SVS-opencpop-cascade.md#4-inference-from-raw-inputs) & [B](docs/README-SVS-opencpop-e2e.md#4-inference-from-raw-inputs); Add Interactive SVS running on [HuggingFace🤗 SVS](https://huggingface.co/spaces/Silentlin/DiffSinger).
+ - Mar.2, 2022: MIDI-B-version.
+ - Mar.1, 2022: [NeuralSVB](https://github.com/MoonInTheRiver/NeuralSVB), for singing voice beautifying, has been released.
+ - Feb.13, 2022: [NATSpeech](https://github.com/NATSpeech/NATSpeech), the improved code framework, which contains the implementations of DiffSpeech and our NeurIPS-2021 work [PortaSpeech](https://openreview.net/forum?id=xmJsuh8xlq) has been released. 
+ - Jan.29, 2022: support MIDI-A-version SVS.
+ - Jan.13, 2022: support SVS, release PopCS dataset.
+ - Dec.19, 2021: support TTS. [HuggingFace🤗 TTS](https://huggingface.co/spaces/NATSpeech/DiffSpeech)
+ 
+:rocket: **News**: 
+ - Feb.24, 2022: Our new work, NeuralSVB was accepted by ACL-2022 [![arXiv](https://img.shields.io/badge/arXiv-Paper-<COLOR>.svg)](https://arxiv.org/abs/2202.13277). [Demo Page](https://neuralsvb.github.io).
+ - Dec.01, 2021: DiffSinger was accepted by AAAI-2022.
+ - Sep.29, 2021: Our recent work `PortaSpeech: Portable and High-Quality Generative Text-to-Speech` was accepted by NeurIPS-2021 [![arXiv](https://img.shields.io/badge/arXiv-Paper-<COLOR>.svg)](https://arxiv.org/abs/2109.15166) .
+ - May.06, 2021: We submitted DiffSinger to Arxiv [![arXiv](https://img.shields.io/badge/arXiv-Paper-<COLOR>.svg)](https://arxiv.org/abs/2105.02446).
+
+## Environments
+1. If you want to use env of anaconda:
+    ```sh
+    conda create -n your_env_name python=3.8
+    source activate your_env_name 
+    pip install -r requirements_2080.txt   (GPU 2080Ti, CUDA 10.2)
+    or pip install -r requirements_3090.txt   (GPU 3090, CUDA 11.4)
+    ```
+
+2. Or, if you want to use virtual env of python:
+    ```sh
+    ## Install Python 3.8 first. 
+    python -m venv venv
+    source venv/bin/activate
+    # install requirements.
+    pip install -U pip
+    pip install Cython numpy==1.19.1
+    pip install torch==1.9.0
+    pip install -r requirements.txt
+    ```
+
+## Documents
+- [Run DiffSpeech (TTS version)](docs/README-TTS.md).
+- [Run DiffSinger (SVS version)](docs/README-SVS.md).
+
+## Overview
+| Mel Pipeline                                                                                | Dataset                                                  | Pitch Input       | F0 Prediction |   Acceleration Method       | Vocoder                       |
+| ------------------------------------------------------------------------------------------- | ---------------------------------------------------------| ----------------- | ------------- | --------------------------- | ----------------------------- |
+| [DiffSpeech (Text->F0, Text+F0->Mel, Mel->Wav)](docs/README-TTS.md)                         | [Ljspeech](https://keithito.com/LJ-Speech-Dataset/)      | None              | Explicit      | Shallow Diffusion           | NSF-HiFiGAN                   |
+| [DiffSinger (Lyric+F0->Mel, Mel->Wav)](docs/README-SVS-popcs.md)                            | [PopCS](https://github.com/MoonInTheRiver/DiffSinger)    | Ground-Truth F0   | None          | Shallow Diffusion           | NSF-HiFiGAN                   |
+| [DiffSinger (Lyric+MIDI->F0, Lyric+F0->Mel, Mel->Wav)](docs/README-SVS-opencpop-cascade.md) | [OpenCpop](https://wenet.org.cn/opencpop/)               | MIDI              | Explicit      | Shallow Diffusion           | NSF-HiFiGAN                   |
+| [FFT-Singer (Lyric+MIDI->F0, Lyric+F0->Mel, Mel->Wav)](docs/README-SVS-opencpop-cascade.md) | [OpenCpop](https://wenet.org.cn/opencpop/)               | MIDI              | Explicit      | Invalid                     | NSF-HiFiGAN                   |
+| [DiffSinger (Lyric+MIDI->Mel, Mel->Wav)](docs/README-SVS-opencpop-e2e.md)                   | [OpenCpop](https://wenet.org.cn/opencpop/)               | MIDI              | Implicit      | None                        | Pitch-Extractor + NSF-HiFiGAN |
+| [DiffSinger+PNDM (Lyric+MIDI->Mel, Mel->Wav)](docs/README-SVS-opencpop-pndm.md)             | [OpenCpop](https://wenet.org.cn/opencpop/)               | MIDI              | Implicit      | PLMS                        | Pitch-Extractor + NSF-HiFiGAN |
+ 
+
+## Tensorboard
+```sh
+tensorboard --logdir_spec exp_name
+```
+<table style="width:100%">
+  <tr>
+    <td><img src="resources/tfb.png" alt="Tensorboard" height="250"></td>
+  </tr>
+</table>
+
+## Citation
+    @article{liu2021diffsinger,
+      title={Diffsinger: Singing voice synthesis via shallow diffusion mechanism},
+      author={Liu, Jinglin and Li, Chengxi and Ren, Yi and Chen, Feiyang and Liu, Peng and Zhao, Zhou},
+      journal={arXiv preprint arXiv:2105.02446},
+      volume={2},
+      year={2021}}
+
+
+## Acknowledgements
+* lucidrains' [denoising-diffusion-pytorch](https://github.com/lucidrains/denoising-diffusion-pytorch)
+* Official [PyTorch Lightning](https://github.com/PyTorchLightning/pytorch-lightning)
+* kan-bayashi's [ParallelWaveGAN](https://github.com/kan-bayashi/ParallelWaveGAN)
+* jik876's [HifiGAN](https://github.com/jik876/hifi-gan)
+* Official [espnet](https://github.com/espnet/espnet)
+* lmnt-com's [DiffWave](https://github.com/lmnt-com/diffwave)
+* keonlee9420's [Implementation](https://github.com/keonlee9420/DiffSinger). 
+
+Especially thanks to:
+
+* Team Openvpi's maintenance: [DiffSinger](https://github.com/openvpi/DiffSinger).
+* Your re-creation and sharing.
+    
\ No newline at end of file

歌声合成/configs/config_base.yaml

+# task
+binary_data_dir: ''
+work_dir: '' # experiment directory.
+infer: false # infer
+seed: 1234
+debug: false
+save_codes:
+  - configs
+  - modules
+  - tasks
+  - utils
+  - usr
+
+#############
+# dataset
+#############
+ds_workers: 1
+test_num: 100
+valid_num: 100
+endless_ds: false
+sort_by_len: true
+
+#########
+# train and eval
+#########
+load_ckpt: ''
+save_ckpt: true
+save_best: false
+num_ckpt_keep: 3
+clip_grad_norm: 0
+accumulate_grad_batches: 1
+log_interval: 100
+num_sanity_val_steps: 5  # steps of validation at the beginning
+check_val_every_n_epoch: 10
+val_check_interval: 2000
+max_epochs: 1000
+max_updates: 160000
+max_tokens: 31250
+max_sentences: 100000
+max_eval_tokens: -1
+max_eval_sentences: -1
+test_input_dir: ''

歌声合成/configs/singing/base.yaml

+base_config:
+  - configs/tts/base.yaml
+  - configs/tts/base_zh.yaml
+
+
+datasets: []
+test_prefixes: []
+test_num: 0
+valid_num: 0
+
+pre_align_cls: data_gen.singing.pre_align.SingingPreAlign
+binarizer_cls: data_gen.singing.binarize.SingingBinarizer
+pre_align_args:
+  use_tone: false # for ZH
+  forced_align: mfa
+  use_sox: true
+hop_size: 128            # Hop size.
+fft_size: 512           # FFT size.
+win_size: 512           # FFT size.
+max_frames: 8000
+fmin: 50                 # Minimum freq in mel basis calculation.
+fmax: 11025               # Maximum frequency in mel basis calculation.
+pitch_type: frame
+
+hidden_size: 256
+mel_loss: "ssim:0.5|l1:0.5"
+lambda_f0: 0.0
+lambda_uv: 0.0
+lambda_energy: 0.0
+lambda_ph_dur: 0.0
+lambda_sent_dur: 0.0
+lambda_word_dur: 0.0
+predictor_grad: 0.0
+use_spk_embed: true
+use_spk_id: false
+
+max_tokens: 20000
+max_updates: 400000
+num_spk: 100
+save_f0: true
+use_gt_dur: true
+use_gt_f0: true

歌声合成/configs/singing/fs2.yaml

+base_config:
+  - configs/tts/fs2.yaml
+  - configs/singing/base.yaml

歌声合成/configs/tts/base.yaml

+# task
+base_config: configs/config_base.yaml
+task_cls: ''
+#############
+# dataset
+#############
+raw_data_dir: ''
+processed_data_dir: ''
+binary_data_dir: ''
+dict_dir: ''
+pre_align_cls: ''
+binarizer_cls: data_gen.tts.base_binarizer.BaseBinarizer
+pre_align_args:
+  use_tone: true # for ZH
+  forced_align: mfa
+  use_sox: false
+  txt_processor: en
+  allow_no_txt: false
+  denoise: false
+binarization_args:
+  shuffle: false
+  with_txt: true
+  with_wav: false
+  with_align: true
+  with_spk_embed: true
+  with_f0: true
+  with_f0cwt: true
+
+loud_norm: false
+endless_ds: true
+reset_phone_dict: true
+
+test_num: 100
+valid_num: 100
+max_frames: 1550
+max_input_tokens: 1550
+audio_num_mel_bins: 80
+audio_sample_rate: 22050
+hop_size: 256  # For 22050Hz, 275 ~= 12.5 ms (0.0125 * sample_rate)
+win_size: 1024  # For 22050Hz, 1100 ~= 50 ms (If None, win_size: fft_size) (0.05 * sample_rate)
+fmin: 80  # Set this to 55 if your speaker is male! if female, 95 should help taking off noise. (To test depending on dataset. Pitch info: male~[65, 260], female~[100, 525])
+fmax: 7600  # To be increased/reduced depending on data.
+fft_size: 1024  # Extra window size is filled with 0 paddings to match this parameter
+min_level_db: -100
+num_spk: 1
+mel_vmin: -6
+mel_vmax: 1.5
+ds_workers: 4
+
+#########
+# model
+#########
+dropout: 0.1
+enc_layers: 4
+dec_layers: 4
+hidden_size: 384
+num_heads: 2
+prenet_dropout: 0.5
+prenet_hidden_size: 256
+stop_token_weight: 5.0
+enc_ffn_kernel_size: 9
+dec_ffn_kernel_size: 9
+ffn_act: gelu
+ffn_padding: 'SAME'
+
+
+###########
+# optimization
+###########
+lr: 2.0
+warmup_updates: 8000
+optimizer_adam_beta1: 0.9
+optimizer_adam_beta2: 0.98
+weight_decay: 0
+clip_grad_norm: 1
+
+
+###########
+# train and eval
+###########
+max_tokens: 30000
+max_sentences: 100000
+max_eval_sentences: 1
+max_eval_tokens: 60000
+train_set_name: 'train'
+valid_set_name: 'valid'
+test_set_name: 'test'
+vocoder: pwg
+vocoder_ckpt: ''
+profile_infer: false
+out_wav_norm: false
+save_gt: false
+save_f0: false
+gen_dir_name: ''
+use_denoise: false

歌声合成/configs/tts/base_zh.yaml

+pre_align_args:
+  txt_processor: zh_g2pM
+binarizer_cls: data_gen.tts.binarizer_zh.ZhBinarizer
\ No newline at end of file

歌声合成/configs/tts/fs2.yaml

+base_config: configs/tts/base.yaml
+task_cls: tasks.tts.fs2.FastSpeech2Task
+
+# model
+hidden_size: 256
+dropout: 0.1
+encoder_type: fft # fft|tacotron|tacotron2|conformer
+encoder_K: 8 # for tacotron encoder
+decoder_type: fft # fft|rnn|conv|conformer
+use_pos_embed: true
+
+# duration
+predictor_hidden: -1
+predictor_kernel: 5
+predictor_layers: 2
+dur_predictor_kernel: 3
+dur_predictor_layers: 2
+predictor_dropout: 0.5
+
+# pitch and energy
+use_pitch_embed: true
+pitch_type: ph # frame|ph|cwt
+use_uv: true
+cwt_hidden_size: 128
+cwt_layers: 2
+cwt_loss: l1
+cwt_add_f0_loss: false
+cwt_std_scale: 0.8
+
+pitch_ar: false
+#pitch_embed_type: 0q
+pitch_loss: 'l1' # l1|l2|ssim
+pitch_norm: log
+use_energy_embed: false
+
+# reference encoder and speaker embedding
+use_spk_id: false
+use_split_spk_id: false
+use_spk_embed: false
+use_var_enc: false
+lambda_commit: 0.25
+ref_norm_layer: bn
+pitch_enc_hidden_stride_kernel:
+  - 0,2,5 # conv_hidden_size, conv_stride, conv_kernel_size. conv_hidden_size=0: use hidden_size
+  - 0,2,5
+  - 0,2,5
+dur_enc_hidden_stride_kernel:
+  - 0,2,3 # conv_hidden_size, conv_stride, conv_kernel_size. conv_hidden_size=0: use hidden_size
+  - 0,2,3
+  - 0,1,3
+
+
+# mel
+mel_loss: l1:0.5|ssim:0.5 # l1|l2|gdl|ssim or l1:0.5|ssim:0.5
+
+# loss lambda
+lambda_f0: 1.0
+lambda_uv: 1.0
+lambda_energy: 0.1
+lambda_ph_dur: 1.0
+lambda_sent_dur: 1.0
+lambda_word_dur: 1.0
+predictor_grad: 0.1
+
+# train and eval
+pretrain_fs_ckpt: ''
+warmup_updates: 2000
+max_tokens: 32000
+max_sentences: 100000
+max_eval_sentences: 1
+max_updates: 120000
+num_valid_plots: 5
+num_test_samples: 0
+test_ids: []
+use_gt_dur: false
+use_gt_f0: false
+
+# exp
+dur_loss: mse # huber|mol
+norm_type: gn
\ No newline at end of file

歌声合成/configs/tts/hifigan.yaml

+base_config: configs/tts/pwg.yaml
+task_cls: tasks.vocoder.hifigan.HifiGanTask
+resblock: "1"
+adam_b1: 0.8
+adam_b2: 0.99
+upsample_rates: [ 8,8,2,2 ]
+upsample_kernel_sizes: [ 16,16,4,4 ]
+upsample_initial_channel: 128
+resblock_kernel_sizes: [ 3,7,11 ]
+resblock_dilation_sizes: [ [ 1,3,5 ], [ 1,3,5 ], [ 1,3,5 ] ]
+
+lambda_mel: 45.0
+
+max_samples: 8192
+max_sentences: 16
+
+generator_params:
+  lr: 0.0002            # Generator's learning rate.
+  aux_context_window: 0 # Context window size for auxiliary feature.
+discriminator_optimizer_params:
+  lr: 0.0002            # Discriminator's learning rate.
\ No newline at end of file

歌声合成/configs/tts/lj/base_mel2wav.yaml

+raw_data_dir: 'data/raw/LJSpeech-1.1'
+processed_data_dir: 'data/processed/ljspeech'
+binary_data_dir: 'data/binary/ljspeech_wav'

歌声合成/configs/tts/lj/base_text2mel.yaml

+raw_data_dir: 'data/raw/LJSpeech-1.1'
+processed_data_dir: 'data/processed/ljspeech'
+binary_data_dir: 'data/binary/ljspeech'
+pre_align_cls: data_gen.tts.lj.pre_align.LJPreAlign
+
+pitch_type: cwt
+mel_loss: l1
+num_test_samples: 20
+test_ids: [ 68, 70, 74, 87, 110, 172, 190, 215, 231, 294,
+            316, 324, 402, 422, 485, 500, 505, 508, 509, 519 ]
+use_energy_embed: false
+test_num: 523
+valid_num: 348
\ No newline at end of file

歌声合成/configs/tts/lj/fs2.yaml

+base_config:
+  - configs/tts/fs2.yaml
+  - configs/tts/lj/base_text2mel.yaml
\ No newline at end of file

歌声合成/configs/tts/lj/hifigan.yaml

+base_config:
+  - configs/tts/hifigan.yaml
+  - configs/tts/lj/base_mel2wav.yaml
\ No newline at end of file

歌声合成/configs/tts/lj/pwg.yaml

+base_config:
+  - configs/tts/pwg.yaml
+  - configs/tts/lj/base_mel2wav.yaml
\ No newline at end of file

歌声合成/configs/tts/pwg.yaml

+base_config: configs/tts/base.yaml
+task_cls: tasks.vocoder.pwg.PwgTask
+
+binarization_args:
+  with_wav: true
+  with_spk_embed: false
+  with_align: false
+test_input_dir: ''
+
+###########
+# train and eval
+###########
+max_samples: 25600
+max_sentences: 5
+max_eval_sentences: 1
+max_updates: 1000000
+val_check_interval: 2000
+
+
+###########################################################
+#                FEATURE EXTRACTION SETTING               #
+###########################################################
+sampling_rate: 22050     # Sampling rate.
+fft_size: 1024           # FFT size.
+hop_size: 256            # Hop size.
+win_length: null         # Window length.
+# If set to null, it will be the same as fft_size.
+window: "hann"           # Window function.
+num_mels: 80             # Number of mel basis.
+fmin: 80                 # Minimum freq in mel basis calculation.
+fmax: 7600               # Maximum frequency in mel basis calculation.
+format: "hdf5"           # Feature file format. "npy" or "hdf5" is supported.
+
+###########################################################
+#         GENERATOR NETWORK ARCHITECTURE SETTING          #
+###########################################################
+generator_params:
+  in_channels: 1        # Number of input channels.
+  out_channels: 1       # Number of output channels.
+  kernel_size: 3        # Kernel size of dilated convolution.
+  layers: 30            # Number of residual block layers.
+  stacks: 3             # Number of stacks i.e., dilation cycles.
+  residual_channels: 64 # Number of channels in residual conv.
+  gate_channels: 128    # Number of channels in gated conv.
+  skip_channels: 64     # Number of channels in skip conv.
+  aux_channels: 80      # Number of channels for auxiliary feature conv.
+  # Must be the same as num_mels.
+  aux_context_window: 2 # Context window size for auxiliary feature.
+  # If set to 2, previous 2 and future 2 frames will be considered.
+  dropout: 0.0          # Dropout rate. 0.0 means no dropout applied.
+  use_weight_norm: true # Whether to use weight norm.
+  # If set to true, it will be applied to all of the conv layers.
+  upsample_net: "ConvInUpsampleNetwork" # Upsampling network architecture.
+  upsample_params:                      # Upsampling network parameters.
+    upsample_scales: [4, 4, 4, 4]     # Upsampling scales. Prodcut of these must be the same as hop size.
+  use_pitch_embed: false
+
+###########################################################
+#       DISCRIMINATOR NETWORK ARCHITECTURE SETTING        #
+###########################################################
+discriminator_params:
+  in_channels: 1        # Number of input channels.
+  out_channels: 1       # Number of output channels.
+  kernel_size: 3        # Number of output channels.
+  layers: 10            # Number of conv layers.
+  conv_channels: 64     # Number of chnn layers.
+  bias: true            # Whether to use bias parameter in conv.
+  use_weight_norm: true # Whether to use weight norm.
+  # If set to true, it will be applied to all of the conv layers.
+  nonlinear_activation: "LeakyReLU" # Nonlinear function after each conv.
+  nonlinear_activation_params:      # Nonlinear function parameters
+    negative_slope: 0.2           # Alpha in LeakyReLU.
+
+###########################################################
+#                   STFT LOSS SETTING                     #
+###########################################################
+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
+use_mel_loss: false
+
+###########################################################
+#               ADVERSARIAL LOSS SETTING                  #
+###########################################################
+lambda_adv: 4.0  # Loss balancing coefficient.
+
+###########################################################
+#             OPTIMIZER & SCHEDULER SETTING               #
+###########################################################
+generator_optimizer_params:
+  lr: 0.0001             # Generator's learning rate.
+  eps: 1.0e-6            # Generator's epsilon.
+  weight_decay: 0.0      # Generator's weight decay coefficient.
+generator_scheduler_params:
+  step_size: 200000      # Generator's scheduler step size.
+  gamma: 0.5             # Generator's scheduler gamma.
+  # At each step size, lr will be multiplied by this parameter.
+generator_grad_norm: 10    # Generator's gradient norm.
+discriminator_optimizer_params:
+  lr: 0.00005            # Discriminator's learning rate.
+  eps: 1.0e-6            # Discriminator's epsilon.
+  weight_decay: 0.0      # Discriminator's weight decay coefficient.
+discriminator_scheduler_params:
+  step_size: 200000      # Discriminator's scheduler step size.
+  gamma: 0.5             # Discriminator's scheduler gamma.
+  # At each step size, lr will be multiplied by this parameter.
+discriminator_grad_norm: 1 # Discriminator's gradient norm.
+disc_start_steps: 40000 # Number of steps to start to train discriminator.

歌声合成/data/processed/ljspeech/dict.txt

+! !
+, ,
+. .
+; ;
+<BOS> <BOS>
+<EOS> <EOS>
+? ?
+AA0 AA0
+AA1 AA1
+AA2 AA2
+AE0 AE0
+AE1 AE1
+AE2 AE2
+AH0 AH0
+AH1 AH1
+AH2 AH2
+AO0 AO0
+AO1 AO1
+AO2 AO2
+AW0 AW0
+AW1 AW1
+AW2 AW2
+AY0 AY0
+AY1 AY1
+AY2 AY2
+B B
+CH CH
+D D
+DH DH
+EH0 EH0
+EH1 EH1
+EH2 EH2
+ER0 ER0
+ER1 ER1
+ER2 ER2
+EY0 EY0
+EY1 EY1
+EY2 EY2
+F F
+G G
+HH HH
+IH0 IH0
+IH1 IH1
+IH2 IH2
+IY0 IY0
+IY1 IY1
+IY2 IY2
+JH JH
+K K
+L L
+M M
+N N
+NG NG
+OW0 OW0
+OW1 OW1
+OW2 OW2
+OY0 OY0
+OY1 OY1
+OY2 OY2
+P P
+R R
+S S
+SH SH
+T T
+TH TH
+UH0 UH0
+UH1 UH1
+UH2 UH2
+UW0 UW0
+UW1 UW1
+UW2 UW2
+V V
+W W
+Y Y
+Z Z
+ZH ZH
+| |