Maybe Wraps layer with Weight Normalization wrapper. Args; layer: tf layers instance, the layer candidate for normalization weight_normalization

Applies multihead attention. Args: queries: A 3d tensor with shape of [N, T_q, C_q]. keys: A 3d tensor with shape of [N, T_k, C_k].

Stop on EOS. Otherwise, pass the last output as the next input and pass through state.

Compute number of time frames of spectrogram

整型数字转汉字

p1 | 前前基元 p2 | 前一基元 p3 | 当前基元 p4 | 后一基元 p5 | 后后基元 p6 | 当前音节的元音

format pinyin to mtts's format

Load Tacotron-2 training state from disk. (To skip if not first run)

Reads a TextGrid file into a dictionary object each dictionary has the following keys: "start" "stop" "name" "tier" Poin

Compute receptive field size. Args: total_layers; int num_cycles: int kernel_size: int dilation: callable, function used to compute dilation

seprate syllable to consonant + ' ' + vowel

Converts a sequence of IDs back to a string

Return list of dims, statically where possible.

Forward step Args: x: Tensor of shape [batch_size, channels, time_length], One-hot encoded audio signal. c: Tensor of shape [batch_size, cin_

发音时长

Writes a list of textgrid dictionaries to a csv file. If no filename is specified, csv is printed to standard out.

Return an initial (zero) state tuple for this `AttentionWrapper`. Args: batch_size: `0D` integer tensor: the batch size.

Runs vanilla LSTM Cell and applies zoneout.

Score the query based on the keys and values. Args: query: Tensor of dtype matching `self.values` and shape `[batch_size, query_depth]`. s

Initialize decoder parameters Args: prenet: A tensorflow fully connected layer acting as the decoder pre-net attentio

Initialize CustomDecoder. Args: cell: An `RNNCell` instance. helper: A `Helper` instance. initial_state: A (possibly nested tuple of...) te

Initializer with possibility to set different zoneout values for cell/hidden states.

Args: is_training: Boolean, determines if the model is training or in inference to control dropout kernel_size: tuple

Args: is_training: Boolean, determines if the model is training or in inference to control zoneout size: integer, the