Calculate hinge loss for fake samples.

Calculate hinge loss for real samples.

Calculate MSE loss for generator.

Calculate MSE loss for fake samples.

Calculate MSE loss for real samples.

Decompose the input signal into num_split subscale signals. Args: x (Tensor): Input signal with shape (B, 1, T).

Decompose the input signal into num_split subscale signals. Args: x (Tensor): Input signal with shape (B, 1, T).

Checks if a specified dataset exists in an HDF5 file. Args: hdf5_name (str): HDF5 file path. hdf5_path (str): Dataset path w

Dynamically imports a Python class from a module using its full module path. Args: module_class (str): Full module path in the forma

Extract various acoustic features (F0, mel-spectrogram, spectral envelope, etc.) from a list of WAV files. Args: queue (multiprocess

Replaces zero values in the input array with the nearest non-zero values from neighboring indices. Args: arr (ndarray): Input array.

Calculate forward propagation. Args: real (Tensor): Real part of the input tensor with shape (batch, channels, height, w

Applies the activation function to the real and imaginary parts of the input complex tensor. Args: real (Tensor): The re

Apply batch normalization to the input complex-valued tensor. Args: real (Tensor): Real part of the input tensor with sh

Apply layer normalization to the input complex tensor. Args: real (Tensor): The real part of the input tensor with shape

Apply batch normalization to the input complex-valued tensor. Args: real (Tensor): Real part of the input tensor with sh

Apply layer normalization to the input complex-valued tensor. Args: real (Tensor): Real part of the input tensor with sh

Calculate forward propagation. Args: x (Tensor): Input tensor with shape (batch, channels, height, width). Retu

Calculate forward propagation. Args: real (Tensor): Input real part tensor with shape (batch, channels, height, width).

Perform the forward Short-Time Fourier Transform (STFT) on the input waveform. Args: x (Tensor): Input waveform with sha

Compute mel-spectrogram from the input waveforms. Args: audio (Tensor): Input waveforms with shape (batch, samples) or (

Apply batch normalization to the input tensor. Args: x (Tensor): Input tensor with shape (batch, channels, height, width

Apply normalization to the input tensor. Args: x (Tensor): Input tensor with shape (batch, channels, height, width).

Apply batch normalization to the input tensor. Args: x (Tensor): Input tensor with shape (batch, channels, height, width

Calculate forward propagation. Args: x (Tensor): Input tensor with shape (batch, in_channels, length). d (Te

Calculate forward propagation. Args: x (Tensor): Input tensor with shape (batch, channels, bins, frames). dh

Calculate forward propagation. Args: x (Tensor): Input waveforms with shape (batch, 1, length). Returns:

Calculate forward propagation. Args: x (Tensor): Input waveforms with shape (batch, 1, length). Returns:

Calculate forward propagation. Args: x (Tensor): Input waveforms with shape (batch, 1, length). Returns:

Calculate forward propagation. Args: x (Tensor): Input waveforms with shape (batch, 1, length). Returns:

Calculate forward propagation. Args: x (Tensor): Input waveforms with shape (batch, 1, length). Returns:

Calculate forward propagation. Args: cond (Tensor): Conditioning features with shape (batch, in_channels, frames).

Calculate forward propagation. Args: cond (Tensor): Conditioning features with shape (batch, in_channels, frames).

Calculate forward propagation. Args: cond (Tensor): Conditioning features with shape (batch, in_channels, frames).

Calculate the L1 loss between mel-spectrograms of the generated and target waveforms. Args: x (Tensor): Generated audio

Calculate adversarial loss for both generator and discriminator. Args: p_fakes (List[Tensor]): List of discriminator out

Calculate feature matching loss. Args: fmaps_fake (List[Tensor]): List of discriminator's latent features from generated

Generate Gaussian noise waveforms of specified duration based on input F0 sequences. Args: f0 (Tensor): F0 sequences with shape (bat

Generate pseudo-constant-power harmonic waveforms based on input F0 sequences. The spectral envelope of harmonics is designed to have flat sp

An optimized O(1) generator for Pseudo-Constant-Power Harmonic waveforms. Uses the Dirichlet kernel closed-form identity formula for speed an

Generate pseudo-constant-power harmonic waveforms based on input F0 sequences. The spectral envelope of harmonics is designed to linearly dec

Generate sawtooth waveforms based on F0 sequences. Args: f0 (Tensor): F0 sequences with shape (batch, 1, frames). hop_length

Generate sine waveforms based on F0 sequences. Args: f0 (Tensor): F0 sequences with shape (batch, 1, frames). hop_length (in

Perform the Griffin-Lim algorithm for phase recovery from a magnitude spectrogram. Args: spectrogram (Tensor): Input complex spectro

Initialize weights of the module. Args: m (Any): Module to initialize.

Initialize weights of the module. Args: m (Any): Module to initialize.

Normalizes the loudness of an input monaural audio signal. Args: audio (ndarray): Input audio waveform. sample_rate (int): S

Remove weight normalization from the given module if it has weight normalization applied. Args: m (Any): Module to remove weight nor

Apply spectral normalization to the given module if it is a supported layer type. Args: m (Any): Module to apply spectral normalizat

Divide a list of audio files based on the speaker and prepare speaker-specific configurations. Args: file_list (List[str]): List of

Reconstruct the full-scale signal from subscale signals. Args: xs (List[Tensor]): List of subscale signals, each with sh

Reconstruct the full-scale signal from subscale signals. Args: xs (List[Tensor]): List of subscale signals, each with sh

Apply weight normalization to the given module if it is a supported layer type. Args: m (Any): Module to apply weight normalization