paddlespeech.t2s.datasets.vocoder_batch_fn module

class paddlespeech.t2s.datasets.vocoder_batch_fn.Clip(batch_max_steps=20480, hop_size=256, aux_context_window=0)

Bases: object

Collate functor for training vocoders.

Methods

__call__(batch)

Convert into batch tensors.

class paddlespeech.t2s.datasets.vocoder_batch_fn.Clip_static(batch_max_steps=20480, hop_size=256, aux_context_window=0)

Bases: Clip

Collate functor for training vocoders.

Methods

__call__(batch)

Convert into batch tensors.

class paddlespeech.t2s.datasets.vocoder_batch_fn.WaveRNNClip(mode: str = 'RAW', batch_max_steps: int = 4500, hop_size: int = 300, aux_context_window: int = 2, bits: int = 9, mu_law: bool = True)

Bases: Clip

Methods

__call__(batch)

Convert into batch tensors. Args: batch (list): list of tuple of the pair of audio and features. Audio shape (T, ), features shape(T', C).

to_quant

to_quant(wav)