Read hdf5 dataset. Args: hdf5_name (str): Filename of hdf5 file. hdf5_path (str): Dataset name in hdf5 file. Return:

Run one step.

convolve each column in data_matrix with kernel 类似CNN的那种1d卷积

Return the keys of the scp file.

Write dataset to hdf5. Args: hdf5_name (str): Hdf5 dataset filename. hdf5_path (str): Dataset path in hdf5. write_data (n

Find files recursively. Args: root_dir (str): Root root_dir to find. query (str): Query to find. include_root_dir (bool):

Initialize Conv1d module.

Initialize Conv2d module.

Initialize STFT loss module.

Computes where to split an utterance waveform and its corresponding mel spectrogram to obtain partial utterances of <partial_utterance_n_fra

Computes embeddings for a batch of mel spectrogram. :param frames_batch: a batch mel of spectrogram as a numpy array of float32 of shape

Remove weight normalization module from all of the layers.

Initialize TFResidualStack module. Args: kernel_size (int): Kernel size. channles (int): Number of channels.

librosa implementation of Griffin-Lim Based on https://github.com/librosa/librosa/issues/434

Write to tensorboard.

Computes embeddings for a batch of mel spectrogram. :param frames_batch: a batch mel of spectrogram as Tensor (batch_size, n_frames, n_c

Computes the embeddings of a batch of utterance spectrograms. :param utterances: batch of mel-scale filterbanks of same dura

Load checkpoint. Args: checkpoint_path (str): Checkpoint path to be loaded. load_only_params (bool): Whether to load

Computes the softmax loss according the section 2.1 of GE2E. :param embeds: the embeddings as a tensor of shape (speakers_pe

Low pass filter Args: x (ndarray): Waveform sequence fs (int): Sampling frequency cutoff (float): Cutoff frequency of low

Save checkpoint. Args: checkpoint_path (str): Checkpoint path to be saved.

Perform STFT and convert to magnitude spectrogram. Args: x (Tensor): Input signal tensor (B, T). fft_size (int): FFT size.

Synthesis with PQMF. Args: x (Tensor): Input tensor (B, subbands, T // subbands). Returns: Tensor: Output te

Initialize CausalConv1d module.

Initialize Unconditional Discriminator module. Args: in_channels (int): Number of input channels. out_channels (int):

Initialize Generator module. Args: in_channels (int): Number of input channels. out_channels (int): Number of output

Evaluate model one epoch.

Evaluate model one step.

Generate and save intermediate result.

generate vuv feature by interpolating lf0

Train model one epoch.

Train model one step.

Analysis with PQMF. Args: x (Tensor): Input tensor (B, 1, T). Returns: Tensor: Output tensor (B, subbands, T

Apply weight normalization module from all of the layers.

Apply weight normalization module from all of the layers.

Apply weight normalization module from all of the layers.

Convert F0 to continuous F0 Args: f0 (ndarray): original f0 sequence with the shape (T) Return: (ndarray): continuous f0 with

Design prototype filter for PQMF. This method is based on `A Kaiser window approach for the design of prototype filters of cosine modulated f

Computes embeddings for a batch of mel spectrogram. :param frames_batch: a batch mel of spectrogram as Tensor (batch_size, n_frames, n_c

Computes an embedding for a single utterance. # TODO: handle multiple wavs to benefit from batching on GPU :param wav: a preprocessed (s

Calculate forward propagation. Args: x (Tensor): Input noise signal (B, 1, T). c (Tensor): Local conditioning auxilia

Computes the embeddings of a batch of utterance spectrograms. :param utterances: batch of mel-scale filterbanks of same duration as

Perform inference. Args: c (Union[Tensor, ndarray]): Local conditioning auxiliary features (T' ,C). x (Union[Tensor,

Return the keys of the scp file.

Load trained model. Args: checkpoint (str): Checkpoint path. config (dict): Configuration dict. Return: torch.nn.Mod

Run training process.

Run preprocessing process.

Run decoding process.

Launch distributed processes.

Compute number of time frames of spectrogram

Parse arguments.

Samples a batch of <count> unique partial utterances from the disk in a way that all utterances come up at least once every two cycl