PaddleSpeech/examples/other/tts_finetune/tts3

Finetune your own AM based on FastSpeech2 with AISHELL-3.

This example shows how to finetune your own AM based on FastSpeech2 with AISHELL-3. We use part of csmsc's data (top 200) as finetune data in this example. The example is implemented according to this discussion. Thanks to the developer for the idea.

We use AISHELL-3 to train a multi-speaker fastspeech2 model. You can refer examples/aishell3/tts3 to train multi-speaker fastspeech2 from scratch.

Prepare

Download Pretrained Fastspeech2 model

Assume the path to the model is ./pretrained_models. Download pretrained fastspeech2 model with aishell3: fastspeech2_aishell3_ckpt_1.1.0.zip.

mkdir -p pretrained_models && cd pretrained_models
wget https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_aishell3_ckpt_1.1.0.zip 
unzip fastspeech2_aishell3_ckpt_1.1.0.zip
cd ../

Download MFA tools and pretrained model

Assume the path to the MFA tool is ./tools. Download MFA and pretrained MFA models with aishell3: aishell3_model.zip.

mkdir -p tools && cd tools
# mfa tool
wget https://github.com/MontrealCorpusTools/Montreal-Forced-Aligner/releases/download/v1.0.1/montreal-forced-aligner_linux.tar.gz
tar xvf montreal-forced-aligner_linux.tar.gz
cp montreal-forced-aligner/lib/libpython3.6m.so.1.0 montreal-forced-aligner/lib/libpython3.6m.so
# pretrained mfa model
mkdir -p aligner && cd aligner
wget https://paddlespeech.bj.bcebos.com/MFA/ernie_sat/aishell3_model.zip
unzip aishell3_model.zip
wget https://paddlespeech.bj.bcebos.com/MFA/AISHELL-3/with_tone/simple.lexicon
cd ../../

Prepare your data

Assume the path to the dataset is ./input. This directory contains audio files (*.wav) and label file (labels.txt). The audio file is in wav format. The format of the label file is: utt_id|pinyin. Here is an example of the first 200 data of csmsc.

mkdir -p input && cd input
wget https://paddlespeech.bj.bcebos.com/datasets/csmsc_mini.zip
unzip csmsc_mini.zip
cd ../

When "Prepare" done. The structure of the current directory is listed below.

├── input
│   ├── csmsc_mini
│   │   ├── 000001.wav
│   │   ├── 000002.wav
│   │   ├── 000003.wav
│   │   ├── ...
│   │   ├── 000200.wav
│   │   ├── labels.txt
│   └── csmsc_mini.zip
├── pretrained_models
│   ├── fastspeech2_aishell3_ckpt_1.1.0
│   │   ├── default.yaml
│   │   ├── energy_stats.npy
│   │   ├── phone_id_map.txt
│   │   ├── pitch_stats.npy
│   │   ├── snapshot_iter_96400.pdz
│   │   ├── speaker_id_map.txt
│   │   └── speech_stats.npy
│   └── fastspeech2_aishell3_ckpt_1.1.0.zip
└── tools
    ├── aligner
    │   ├── aishell3_model
    │   ├── aishell3_model.zip
    │   └── simple.lexicon
    ├── montreal-forced-aligner
    │   ├── bin
    │   ├── lib
    │   └── pretrained_models
    └── montreal-forced-aligner_linux.tar.gz
    ...

Set finetune.yaml

finetune.yaml contains some configurations for fine-tuning. You can try various options to fine better result. Arguments:

• batch_size: finetune batch size. Default: -1, means 64 which same to pretrained model
• learning_rate: learning rate. Default: 0.0001
• num_snapshots: number of save models. Default: -1, means 5 which same to pretrained model
• frozen_layers: frozen layers. must be a list. If you don't want to frozen any layer, set [].

Get Started

Run the command below to

1. source path.
2. finetune the model.
3. synthesize wavs.
   • synthesize waveform from text file.

./run.sh

You can choose a range of stages you want to run, or set stage equal to stop-stage to run only one stage.

Model Finetune

Finetune a FastSpeech2 model.

./run.sh --stage 0 --stop-stage 0

stage 0 of run.sh calls finetune.py, here's the complete help message.

usage: finetune.py [-h] [--input_dir INPUT_DIR] [--pretrained_model_dir PRETRAINED_MODEL_DIR]
                [--mfa_dir MFA_DIR] [--dump_dir DUMP_DIR]
                [--output_dir OUTPUT_DIR] [--lang LANG]
                [--ngpu NGPU]

optional arguments:
  -h, --help            show this help message and exit
  --input_dir INPUT_DIR       
                        directory containing audio and label file
  --pretrained_model_dir PRETRAINED_MODEL_DIR
                       Path to pretrained model
  --mfa_dir MFA_DIR    directory to save aligned files
  --dump_dir DUMP_DIR
                       directory to save feature files and metadata
  --output_dir OUTPUT_DIR      
                       directory to save finetune model 
  --lang LANG          Choose input audio language, zh or en
  --ngpu NGPU          if ngpu=0, use cpu
  --epoch EPOCH        the epoch of finetune
  --batch_size BATCH_SIZE        
                       the batch size of finetune, default -1 means same as pretrained model

1. --input_dir is the directory containing audio and label file.
2. --pretrained_model_dir is the directory incluing pretrained fastspeech2_aishell3 model.
3. --mfa_dir is the directory to save the results of aligning from pretrained MFA_aishell3 model.
4. --dump_dir is the directory including audio feature and metadata.
5. --output_dir is the directory to save finetune model.
6. --lang is the language of input audio, zh or en.
7. --ngpu is the number of gpu.
8. --epoch is the epoch of finetune.
9. --batch_size is the batch size of finetune.

Synthesizing

We use HiFiGAN as the neural vocoder. Assume the path to the hifigan model is ./pretrained_models. Download the pretrained HiFiGAN model from hifigan_aishell3_ckpt_0.2.0 and unzip it.

cd pretrained_models
wget https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_aishell3_ckpt_0.2.0.zip
unzip hifigan_aishell3_ckpt_0.2.0.zip
cd ../

HiFiGAN checkpoint contains files listed below.

hifigan_aishell3_ckpt_0.2.0
├── default.yaml                   # default config used to train HiFiGAN
├── feats_stats.npy                # statistics used to normalize spectrogram when training HiFiGAN
└── snapshot_iter_2500000.pdz      # generator parameters of HiFiGAN

Modify ckpt in run.sh to the final model in exp/default/checkpoints.

./run.sh --stage 1 --stop-stage 1

stage 1 of run.sh calls ${BIN_DIR}/../synthesize_e2e.py, which can synthesize waveform from text file.

usage: synthesize_e2e.py [-h]
                         [--am {speedyspeech_csmsc,speedyspeech_aishell3,fastspeech2_csmsc,fastspeech2_ljspeech,fastspeech2_aishell3,fastspeech2_vctk,tacotron2_csmsc,tacotron2_ljspeech}]
                         [--am_config AM_CONFIG] [--am_ckpt AM_CKPT]
                         [--am_stat AM_STAT] [--phones_dict PHONES_DICT]
                         [--tones_dict TONES_DICT]
                         [--speaker_dict SPEAKER_DICT] [--spk_id SPK_ID]
                         [--voc {pwgan_csmsc,pwgan_ljspeech,pwgan_aishell3,pwgan_vctk,mb_melgan_csmsc,style_melgan_csmsc,hifigan_csmsc,hifigan_ljspeech,hifigan_aishell3,hifigan_vctk,wavernn_csmsc}]
                         [--voc_config VOC_CONFIG] [--voc_ckpt VOC_CKPT]
                         [--voc_stat VOC_STAT] [--lang LANG]
                         [--inference_dir INFERENCE_DIR] [--ngpu NGPU]
                         [--text TEXT] [--output_dir OUTPUT_DIR]

Synthesize with acoustic model & vocoder

optional arguments:
  -h, --help            show this help message and exit
  --am {speedyspeech_csmsc,speedyspeech_aishell3,fastspeech2_csmsc,fastspeech2_ljspeech,fastspeech2_aishell3,fastspeech2_vctk,tacotron2_csmsc,tacotron2_ljspeech}
                        Choose acoustic model type of tts task.
  --am_config AM_CONFIG
                        Config of acoustic model.
  --am_ckpt AM_CKPT     Checkpoint file of acoustic model.
  --am_stat AM_STAT     mean and standard deviation used to normalize
                        spectrogram when training acoustic model.
  --phones_dict PHONES_DICT
                        phone vocabulary file.
  --tones_dict TONES_DICT
                        tone vocabulary file.
  --speaker_dict SPEAKER_DICT
                        speaker id map file.
  --spk_id SPK_ID       spk id for multi speaker acoustic model
  --voc {pwgan_csmsc,pwgan_ljspeech,pwgan_aishell3,pwgan_vctk,mb_melgan_csmsc,style_melgan_csmsc,hifigan_csmsc,hifigan_ljspeech,hifigan_aishell3,hifigan_vctk,wavernn_csmsc}
                        Choose vocoder type of tts task.
  --voc_config VOC_CONFIG
                        Config of voc.
  --voc_ckpt VOC_CKPT   Checkpoint file of voc.
  --voc_stat VOC_STAT   mean and standard deviation used to normalize
                        spectrogram when training voc.
  --lang LANG           Choose model language. zh or en
  --inference_dir INFERENCE_DIR
                        dir to save inference models
  --ngpu NGPU           if ngpu == 0, use cpu.
  --text TEXT           text to synthesize, a 'utt_id sentence' pair per line.
  --output_dir OUTPUT_DIR
                        output dir.

1. --am is acoustic model type with the format {model_name}_{dataset}
2. --am_config, --am_ckpt, --am_stat, --phones_dict --speaker_dict are arguments for acoustic model, which correspond to the 5 files in the fastspeech2 pretrained model.
3. --voc is vocoder type with the format {model_name}_{dataset}
4. --voc_config, --voc_ckpt, --voc_stat are arguments for vocoder, which correspond to the 3 files in the parallel wavegan pretrained model.
5. --lang is the model language, which can be zh or en.
6. --text is the text file, which contains sentences to synthesize.
7. --output_dir is the directory to save synthesized audio files.
8. --ngpu is the number of gpus to use, if ngpu == 0, use cpu.

Tips

If you want to get better audio quality, you can use more audios to finetune. More finetune results can be found on finetune-fastspeech2-for-csmsc. The results show the effect on csmsc_mini: Freeze encoder > Non Frozen > Freeze encoder && duration_predictor.