PaddleSpeech/examples/vctk/GANVocoder/parallelwave_gan/ljspeech

Parallel WaveGAN with the LJSpeech-1.1 dataset

This example contains code used to train a parallel wavegan model with LJSpeech-1.1.

Preprocess the dataset

Download and Extract the datasaet

Download LJSpeech-1.1 from the official website.

Get MFA results for silence trim

We use MFA results to cut silence in the edge of audio. You can download from here ljspeech_alignment.tar.gz, or train your own MFA model reference to use_mfa example of our repo.

Preprocess the dataset

Assume the path to the dataset is ~/datasets/LJSpeech-1.1. Assume the path to the MFA result of LJSpeech-1.1 is ./ljspeech_alignment. Run the command below to preprocess the dataset.

./preprocess.sh

When it is done. A dump folder is created in the current directory. The structure of the dump folder is listed below.

dump
├── dev
│   ├── norm
│   └── raw
├── test
│   ├── norm
│   └── raw
└── train
    ├── norm
    ├── raw
    └── feats_stats.npy

The dataset is split into 3 parts, namely train, dev and test, each of which contains a norm and raw subfolder. The raw folder contains log magnitude of mel spectrogram of each utterances, while the norm folder contains normalized spectrogram. The statistics used to normalize the spectrogram is computed from the training set, which is located in dump/train/feats_stats.npy.

Also there is a metadata.jsonl in each subfolder. It is a table-like file which contains id and paths to spectrogam of each utterance.

Train the model

./run.sh calls ../train.py.

./run.sh

Here's the complete help message.

usage: train.py [-h] [--config CONFIG] [--train-metadata TRAIN_METADATA]
                     [--dev-metadata DEV_METADATA] [--output-dir OUTPUT_DIR]
                     [--device DEVICE] [--nprocs NPROCS] [--verbose VERBOSE]
                     [--batch-size BATCH_SIZE] [--max-iter MAX_ITER]
                     [--run-benchmark RUN_BENCHMARK]
                     [--profiler_options PROFILER_OPTIONS]

Train a ParallelWaveGAN model.

optional arguments:
  -h, --help            show this help message and exit
  --config CONFIG       config file to overwrite default config.
  --train-metadata TRAIN_METADATA
                        training data.
  --dev-metadata DEV_METADATA
                        dev data.
  --output-dir OUTPUT_DIR
                        output dir.
  --device DEVICE       device type to use.
  --nprocs NPROCS       number of processes.
  --verbose VERBOSE     verbose.

benchmark:
  arguments related to benchmark.

  --batch-size BATCH_SIZE
                        batch size.
  --max-iter MAX_ITER   train max steps.
  --run-benchmark RUN_BENCHMARK
                        runing benchmark or not, if True, use the --batch-size
                        and --max-iter.
  --profiler_options PROFILER_OPTIONS
                        The option of profiler, which should be in format
                        "key1=value1;key2=value2;key3=value3".

1. --config is a config file in yaml format to overwrite the default config, which can be found at conf/default.yaml.
2. --train-metadata and --dev-metadata should be the metadata file in the normalized subfolder of train and dev in the dump folder.
3. --output-dir is the directory to save the results of the experiment. Checkpoints are save in checkpoints/ inside this directory.
4. --device is the type of the device to run the experiment, 'cpu' or 'gpu' are supported.
5. --nprocs is the number of processes to run in parallel, note that nprocs > 1 is only supported when --device is 'gpu'.

Pretrained Models

Pretrained models can be downloaded here:

1. Parallel WaveGAN checkpoint. pwg_ljspeech_ckpt_0.5.zip, which is used as a vocoder in the end-to-end inference script.

Parallel WaveGAN checkpoint contains files listed below.

pwg_ljspeech_ckpt_0.5
├── pwg_default.yaml              # default config used to train parallel wavegan
├── pwg_snapshot_iter_400000.pdz  # generator parameters of parallel wavegan
└── pwg_stats.npy                 # statistics used to normalize spectrogram when training parallel wavegan

Synthesize

synthesize.sh calls ../synthesize.py , which can synthesize waveform from metadata.jsonl.

./synthesize.sh

usage: synthesize.py [-h] [--config CONFIG] [--checkpoint CHECKPOINT]
                     [--test-metadata TEST_METADATA] [--output-dir OUTPUT_DIR]
                     [--device DEVICE] [--verbose VERBOSE]

Synthesize with parallel wavegan.

optional arguments:
  -h, --help            show this help message and exit
  --config CONFIG       parallel wavegan config file.
  --checkpoint CHECKPOINT
                        snapshot to load.
  --test-metadata TEST_METADATA
                        dev data.
  --output-dir OUTPUT_DIR
                        output dir.
  --device DEVICE       device to run.
  --verbose VERBOSE     verbose.

1. --config parallel wavegan config file. You should use the same config with which the model is trained.
2. --checkpoint is the checkpoint to load. Pick one of the checkpoints from checkpoints inside the training output directory. If you use the pretrained model, use the pwg_snapshot_iter_400000.pdz.
3. --test-metadata is the metadata of the test dataset. Use the metadata.jsonl in the dev/norm subfolder from the processed directory.
4. --output-dir is the directory to save the synthesized audio files.
5. --device is the type of device to run synthesis, 'cpu' and 'gpu' are supported.

Acknowledgement

We adapted some code from https://github.com/kan-bayashi/ParallelWaveGAN.