torchaudio.prototype.models.hifigan_vocoder¶
- torchaudio.prototype.models.hifigan_vocoder(in_channels: int, upsample_rates: Tuple[int, ...], upsample_initial_channel: int, upsample_kernel_sizes: Tuple[int, ...], resblock_kernel_sizes: Tuple[int, ...], resblock_dilation_sizes: Tuple[Tuple[int, ...], ...], resblock_type: int, lrelu_slope: float) HiFiGANVocoder[原始碼]¶
建構 HiFi GAN Vocoder [Kong et al., 2020]。
- 參數:
in_channels (int) – 請參閱
HiFiGANVocoder。upsample_rates (tuple of
int) – 請參閱HiFiGANVocoder。upsample_initial_channel (int) – 請參閱
HiFiGANVocoder。upsample_kernel_sizes (tuple of
int) – 請參閱HiFiGANVocoder。resblock_kernel_sizes (tuple of
int) – 請參閱HiFiGANVocoder。resblock_dilation_sizes (tuple of tuples of
int) – 請參閱HiFiGANVocoder。resblock_type (int, 1 或 2) – 請參閱
HiFiGANVocoder。
- 返回:
產生的模型。
- 返回類型:
