| from customs.make_custom_dataset import create_dataset | |
| ''' | |
| How should the data_dir be created? | |
| Place the necessary audio files in data_dir. | |
| Transcription, tokenization, etc. of the audio files are done by the create_dataset function. | |
| data_dir | |
| ├── bpe_69.json | |
| ├── utt1.wav | |
| ├── utt2.wav | |
| ├── utt3.wav | |
| ...... | |
| └── utt{n}.wav | |
| ''' | |
| data_dir = "VALL-E-X_Dataset/data_dir" | |
| create_dataset(data_dir, audio_path, filename, transcription, duration, sample_rate, asr_model, dataloader_process_only=True) |