espnet2.tts.prodiff.denoiser.noise_scheduler

source

espnet2.tts.prodiff.denoiser.noise_scheduler(sched_type: str, timesteps: int, min_beta: float = 0.0, max_beta: float = 0.01, s: float = 0.008) → Tensor

Noise Scheduler.

This function generates a noise schedule based on the specified type of scheduler. The noise schedule can be linear, cosine, or vpsde, and it returns a tensor representing the noise values at each time step.

• Parameters:
  • sched_type (str) – Type of scheduler. Options are “linear”, “cosine”, and “vpsde”.
  • timesteps (int) – Number of time steps.
  • min_beta (float , optional) – Minimum beta. Defaults to 0.0.
  • max_beta (float , optional) – Maximum beta. Defaults to 0.01.
  • s (float , optional) – Scheduler intersection. Defaults to 0.008.
• Returns: Tensor containing noise values for each time step.
• Return type: torch.Tensor
• Raises:NotImplementedError – If an unsupported scheduler type is provided.

Examples

>>> noise = noise_scheduler("linear", 100)
>>> print(noise.shape)
torch.Size([100])

>>> noise = noise_scheduler("cosine", 100)
>>> print(noise.shape)
torch.Size([100])

>>> noise = noise_scheduler("vpsde", 100, min_beta=0.1, max_beta=0.2)
>>> print(noise.shape)
torch.Size([100])