Returns a random number strictly between 0 and 1.

Try apply weightnorm for all layer in list_layers.

Remove outlier from x.

Args: before_outs: the outputs before postnet logits_stack: the logits of all steps Returns:: af

Add a task

Apply optional mask for encoder. Args: xs (tf.Tensor): padded input, (B, L, D), L for max length mask (tf.Tensor): mask for xs,

return tf.data.Dataset object

Apply attention rescoring decoding, CTC prefix beam search is applied first to get nbest, then we resoring the nbest on atten

batch beam search for transformer model Args: samples: the data source to be decoded beam_size: beam size

build in keras layer

reshape from [N T D C] -> [N T D*C] using tf.shape(x), which generate a tensor instead of x.shape

complex addition

compute cmvn file

compute cmvn for filtered entries using kaldi-format data

compuate cmvn of mel-spec,f0 and energy

compute the logit length

compute the logit length

used for get logit length

compute mean */

Args: value_length: the length of value, shape: [batch] max_len: the maximun length Returns: init

evaluate the model

evaluate the model

synthesize using vocoder on dataset

Args: y: the true label, shape: [batch, y_steps, feat_dim] Returns:: y0: zeros will be padded as one step to

Args: encoder_output: encoder outputs, shape: [batch, x_step, eunits] input_length: shape: [batch] Returns::

insert eos in labels, batch size should be larger than 1 assuming 0 as the padding,

load model

generate a postional encoding list

Mel to Hertz conversion */

Override existing hyperparameter values, parsing new values from a json object. Args: values_json: String containing a json objec

Create mask for subsequent steps (size, size) with chunk size, this is for streaming encoder Args: size (int): size of mask

Create mask for subsequent steps (size, size) with chunk size, this is for streaming encoder Args: size (int): size of mask

input_(tensor): input tensor dim(int): dimension indices(list): selected indices list

Args: encoder_output: encoder output (batch, x_steps, eunits). input_length: (batch,) prev_y: one step of

entry point for model training, implements train loop :param jsonfile: json file to read configuration from :param Solver: an abstract c

Compute fbank feat, return num frames