Spaces:
Running
Running
| #!/usr/bin/python3 | |
| # -*- coding: utf-8 -*- | |
| from toolbox.torchaudio.configuration_utils import PretrainedConfig | |
| class DTLNConfig(PretrainedConfig): | |
| def __init__(self, | |
| sample_rate: int = 8000, | |
| fft_size: int = 200, | |
| hop_size: int = 80, | |
| win_type: str = "hann", | |
| encoder_size: int = 256, | |
| min_snr_db: float = -10, | |
| max_snr_db: float = 20, | |
| lr: float = 0.001, | |
| lr_scheduler: str = "CosineAnnealingLR", | |
| lr_scheduler_kwargs: dict = None, | |
| max_epochs: int = 100, | |
| clip_grad_norm: float = 10., | |
| seed: int = 1234, | |
| num_workers: int = 4, | |
| batch_size: int = 4, | |
| eval_steps: int = 25000, | |
| **kwargs | |
| ): | |
| super(DTLNConfig, self).__init__(**kwargs) | |
| # transform | |
| self.sample_rate = sample_rate | |
| self.fft_size = fft_size | |
| self.hop_size = hop_size | |
| self.win_type = win_type | |
| # model params | |
| self.encoder_size = encoder_size | |
| # data snr | |
| self.min_snr_db = min_snr_db | |
| self.max_snr_db = max_snr_db | |
| # train | |
| self.lr = lr | |
| self.lr_scheduler = lr_scheduler | |
| self.lr_scheduler_kwargs = lr_scheduler_kwargs or dict() | |
| self.max_epochs = max_epochs | |
| self.clip_grad_norm = clip_grad_norm | |
| self.seed = seed | |
| self.num_workers = num_workers | |
| self.batch_size = batch_size | |
| self.eval_steps = eval_steps | |
| def main(): | |
| config = DTLNConfig() | |
| config.to_yaml_file("config.yaml") | |
| return | |
| if __name__ == "__main__": | |
| main() | |