get a sample Args: index (int): index of the entries Returns: dict: sample:: {

get a sample Args: index (int): index of the entries Returns: dict: sample:: {

Construct CTC prefix scorer :param torch.Tensor x: input label posterior sequences (B, T, O) B: batch*beam -> batch; O:V_class :param

Reference: to paper "Multiband Excitation Vocoder"

Args: generator(generator): instance of generator which feed data capacity(int): maximum data to prefetch

Create an instance of `HParams` from keyword arguments. The keyword arguments specify name-values pairs for the hyperparameters. The

parameters for warmup learning rate schedule Args: model_dim (int, optional): usually dim of self-attention vector of transformer

parameters for warmup learning rate schedule Args: model_dim (int, optional): usually dim of self-attention vector of transformer

initialization of multihead attention block Args: d_model: dimension of multi-head attention num_heads: number of att