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[源代码]¶
构建 HiFi GAN 声码器 [Kong 等人，2020]。
- 参数:
in_channels (int) – 请参阅
HiFiGANVocoder。upsample_rates (
int的元组) – 请参阅HiFiGANVocoder。upsample_initial_channel (int) – 请参阅
HiFiGANVocoder。upsample_kernel_sizes (
int的元组) – 请参阅HiFiGANVocoder。resblock_kernel_sizes (
int的元组) – 请参阅HiFiGANVocoder。resblock_dilation_sizes (
int的元组的元组) – 请参阅HiFiGANVocoder。resblock_type (int, 1 或 2) – 请参阅
HiFiGANVocoder。
- 返回值:
生成的模型。
- 返回类型:
